NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
crates.io
crates.io, #1404 most downloaded on crates.io
Implements git packs and related data structures
Latest stable
0.77.0
Last release 1 months ago
22 Oct 2026
Ships on a steady schedule
a new release about every 4 weeks
Nearly every release is documented
notes for 60 of 60 stable releases
1 version withdrawn
withdrawn after publishing
4 years old
61 releases · first in 2023
Drag across the ruler to compare, or tap a release to open it.
4 commits contributed to the release over the course of 14 calendar days.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
reject incomplete delta traversals and overlapping offsets <!-- agent --> A Report describes indexed delta components that cannot be reached from any
Your coding agent can read these notes before it upgrades. Set up the MCP server →
<csr-id-781b6061ee8bb1c25e952ea0de57697029c0ff79/> reject incomplete delta traversals and overlapping offsets <!-- agent --> A Report describes indexed delta components that cannot be reached from any base object yet pass delta-tree verification. Require every indexed item to have been inspected before returning success, using the existing object counter. Legitimate forward references remain supported.
Return corruption errors for duplicate or overlapping index offsets and for an entry at or past the pack end, instead of panicking. The reported decoder cycle and allocation problem is handled by the fix.
Three regressions failed before the fix: an unresolved cycle was accepted, while duplicate offsets and an invalid pack end panicked. Coverage includes self and multi-node components with and without a valid root, plus a valid forward delta base.
Git reference: builtin/index-pack.c::conclude_pack() at
d38352cd43ab9745686d697872408bc3249a153f rejects unresolved deltas.
<csr-id-89a12d4a2558627524ae55151c39950e1b92973f/> reject cyclic delta chains and bound chain allocations <!-- agent --> A report describes delta base cycles that never terminate in header lookups and keep allocating during object decoding. Track a constant-space cycle checkpoint in both loops, without relying on the untrusted pack object count or rejecting valid forward references.
Account for heap-backed delta-chain metadata in the existing per-allocation limit and reserve it fallibly, including when delta payloads are empty. The regressions failed before the fix and cover self, multi-entry, mixed REF/OFS cycles, empty deltas, and valid forward chains.
Git reference: t/t5309-pack-delta-cycles.sh at
d38352cd43ab9745686d697872408bc3249a153f rejects cycles while allowing
forward delta bases.
<csr-id-b41b1a5267cbe7ed31b1600391aedc90c39941ee/> map packs read-only on Windows so large ones can be opened
On Windows if you memory map with copyonwrite (which is what map_copy_read_only does) it charges against the commit limit for that. As such, when opening extremely large packs, i.e. 40GB, This can fail with an out of memory error: ERROR_COMMITMENT_LIMIT (os error 1455, "The paging file is too small for this operation to complete")
The fix is to memory map it plainly. Which is otherwise pretty much the same as it is on other operating systems for read-only memory mapping.
<csr-id-5b6522a8c8e5cc1cfe692ba65e87b8a6fad392ce/> validate multi-pack-index references on access <!-- Byron -->
Looked at everything in detail and added error handling in callbacks
that didn't have them. Reduced the scope of this commit to not overlap
with odb-parallelism branch which makes many changes additionally.
Many of the newly added tests are very specific about error handling,
one might call them overkill, but I left them in for good measure.
Further, errors.rs in gix-odb I just skimmed the test titles off
as everything else would be too time consuming - better have them than
not have them I thought, particularly to support odb-paralellism.
<!-- agent --> A report describes object lookup panics caused by unchecked pack IDs and large-offset ordinals in otherwise structurally valid MIDX files.
Validate these references when reading an object entry, before indexing pack arrays or dereferencing the optional LOFF (Large Offsets) table. Bound LOFF ordinals by the chunk itself, not by the remaining file. Keep MIDX loading limited to structural validation: an eager scan of every offset entry adds linear startup work and faults in memory-mapped pages even when a command needs only a handful of objects. Git likewise checks these references on access.
Make pack_id_and_pack_offset_at_index() return gix_error::Result
and iter() yield fallible entries. Propagate corruption errors through
object and header lookup, delta-base resolution, integrity verification,
and CLI entry listing. ID-only iteration remains available without valid
offset references. Preserve literal high-bit 32-bit offsets when LOFF
is absent.
Return Result<Option<_>> from Find::location_by_oid() and
Find::pack_offsets_and_oid(): absence and lookup failure are different
outcomes, and callers must be able to distinguish them. Propagate errors
through forwarding implementations, object counting, pack-entry iterator
construction, and thin-pack base lookup rather than treating corruption
as a missing object or a reason to recompress. Report a pack that becomes
unavailable during construction as an error instead of panicking.
Retain failed index loads separately from missing files so repeated lookups cannot silently turn corruption into absence. Load pending indices before retrying failures, retry all failed slots, and reconcile disk state when refreshing. This keeps independent valid packs usable and permits recovery after repair or removal without endless refresh loops on persistent failures.
Regressions cover boundary and extreme invalid references, lazy loading, valid LOFF ordinals and literal offsets, integrity verification, corruption propagation through object lookup and pack generation, and repeated, shared, concurrent, repaired, and removed index-load failures.
Git reference: d38352cd43ab9745686d697872408bc3249a153f,
midx.c::midx_for_pack() and nth_midxed_offset().
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
0115e90)794eeef)1d7bac7)5b6522a)781b606)89a12d4)9d5d934)4ddbe11)80f4b03)b41b1a5)d2d078f)95043b9)f819565)1c25831)or_error() at public exception boundaries (535672e)5fb3dcf)
</details>expose the pack header size <!-- agent --> The 12-byte pack header length was repeated in readers, offset validation, and tests, leaving the first-ent
<csr-id-e0825df6b3ab83b4f5897002074b0111c9fcb822/> expose the pack header size <!-- agent --> The 12-byte pack header length was repeated in readers, offset validation, and tests, leaving the first-entry boundary expressed as a magic number.
Expose data::header::SIZE, derive it from the fixed header fields, and use
it wherever code needs the first valid pack-entry offset.
<csr-id-e79baac854f818b717d9500373e30c3525cfc137/> reject delta instructions that exceed the result size <!-- Byron -->
rubber stamp, looked at diff
<!-- agent --> Writing to a byte slice may succeed with a short write. Delta application ignored that byte count, so oversized copy and insert instructions silently truncated their output instead of reporting corrupt data.
Split off an output slice of exactly the required size before copying.
This also removes two mappings of I/O failures that slice writes cannot
produce. A regression covers both copy and insert instructions; all
gix-pack tests pass.
<csr-id-0805bb66782176ab2b24fc25b9e7a111d5c185b0/> avoid aborts in remaining capacity hints
<!-- agent -->
A follow-up audit for GHSA-x862-c2wj-4mwr found that the shared exact_vec()
optimization still used infallible reservation at four other call sites,
multi-index writing reserved from index counts infallibly, and doubled thin-pack
size hints could overflow on 32-bit targets.
Treat exact preallocation as best-effort, propagate multi-index reservation failure, and saturate the thin-pack upper bound. Regression tests cover failed exact preallocation and size-hint overflow without making large allocations.
<csr-id-31d1e21f5373338d47dcfe3aea76c57ff626daf3/> make delta-tree allocation fallible <!-- agent --> GHSA-x862-c2wj-4mwr identified that delta-tree capacity derived from pack metadata used an infallible exact reservation. A rejected allocation could therefore terminate the process instead of returning through the existing indexing error path.
Use fallible exact reservations, reject capacity arithmetic overflow, and apply the existing per-allocation limit to both initial tree vectors. The regression covers allocator rejection and explicit zero limits; bundle and index tests verify the propagated error.
Git reference: builtin/index-pack.c reads the advertised object count in
parse_pack_header() and allocates its object tables with CALLOC_ARRAY. This
change retains gitoxide's eager sizing while making failure recoverable and
configurable.
<csr-id-8d73d3c9a8eccaaed1a1190ace5a0998ec991a68/> reject delta bases in the pack header
<!-- agent -->
OFS_DELTA bases must refer to an earlier pack entry, but the shared offset
helper accepted offsets within the 12-byte pack header. This let all new
consumers treat these malformed distances as valid.
Require the computed base offset to start at or beyond the first entry, matching
git index-pack, and cover both the shared boundary and pack-copy error path.
<csr-id-fd318d887dd867c304daea0c3212157d22cf651d/> reject invalid delta offsets in pack consumers
<!-- agent -->
An audit for the same unchecked OFS_DELTA subtraction found two further
pack-data consumers that trusted raw base distances: delta-tree construction
during verification and entry reuse during pack generation.
Use Header::verified_base_pack_offset() in both paths. Verification now
reports corrupt input through its existing invalid-data error, while pack
generation declines to copy the malformed entry and uses its existing
decode/recompress fallback.
This matches Git at f78ce2f7b6df702f93d40b85d6bda92a3f65da79, where pack
readers reject delta base offsets outside the pack prefix.
<csr-id-e000edb0c7c3e9d2eb6bfc4fd7d74af0537b09a3/> reject invalid thin-pack delta offsets
<!-- agent -->
GHSA-mpf5-465h-mr53 identifies a panic while thin-pack base insertion rewrites
later OFS_DELTA entries. Zero or out-of-bounds base distances reached
checked_sub().expect() instead of being treated as malformed input.
Reuse Header::verified_base_pack_offset() and propagate a typed input
error. This matches Git at f78ce2f7b6df702f93d40b85d6bda92a3f65da79, where
builtin/index-pack.c rejects delta base offsets outside the pack prefix.
<csr-id-e390bc88a3de1d5817c779fe10b36b7c7754a8b9/> support version 3 across pack writers <!-- agent --> An audit following GHSA-633h-mqjc-8rwg found four more assertions that rejected version 3 in pack-writing paths even though versions 2 and 3 share their entry encoding.
Remove the stale assertions so thin-pack repair and the general pack-generation pipeline consistently preserve supported input versions. Extend the bundle regression through the thin-pack lookup path that previously still panicked.
Git baseline: pack_version_ok_native() in pack.h at f78ce2f7b6 accepts
versions 2 and 3.
<csr-id-b851c0f144036f8a122a15a2f52475b95802da9a/> accept version 3 in streaming pack input
<!-- agent -->
BytesToEntriesIter asserted that every decoded pack used version 2 even though
the shared header decoder and file reader accept version 3. This made streaming
input disagree with the rest of the crate and with Git.
Remove the contradictory assertion and exercise a complete version 3 fixture through the streaming iterator. This addresses GHSA-633h-mqjc-8rwg without adding another version gate.
Git baseline: pack_version_ok_native() in pack.h at f78ce2f7b6 accepts
versions 2 and 3 with the same entry layout.
<csr-id-f28f7df44c88e4dc6e7892856cfd454dbcbd36bc/> migrate errors to gix-error <!-- Byron --> rubberstamp
<csr-id-4b42e0ce80ae934cae4f102f44c392581758608f/> raise MSRV to Rust 1.88
<!-- agent -->
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
37860b3)6356013)e79baac)9d0329a)daf73b5)f28f7df)4b9ff51)a095334)4b42e0c)7e35849)888677a)ec98152)a938f52)e0825df)0805bb6)31d1e21)8d73d3c)fd318d8)e000edb)2ec4de7)e390bc8)b851c0f)dda600d)
</details>retain additions from all merge parents <!-- agent --> TreeAdditionsComparedToAncestor cleared its collected changes before every parent diff. For mer
<csr-id-4009a9afca003e339d73ef38eaeccd036b32e3dc/> retain additions from all merge parents
<!-- agent -->
TreeAdditionsComparedToAncestor cleared its collected changes before every
parent diff. For merge commits, objects added relative to an earlier parent were
marked seen and then discarded, producing incomplete packs.
Add a regression graph whose merge tree matches its second parent and verify that the blob found only against the first parent is counted. Clear the delegate once before the parent loop so each parent’s additions accumulate.
Expose a read-only gix_pack::testing::Memory adapter for loose in-memory objects.
Use it with gix_odb::memory::Proxy in the regression and deduplicate the
equivalent benchmark adapter.
Git baseline: 15c6308cf7ad276b306aa5b3ababfbdebfb1a917; list-objects.c:add_edge_parents accounts for every merge parent tree during object traversal.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
4009a9a)3ebca8b)4fa211d)6704303)
</details>3 commits contributed to the release.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
expose index-less parallel pack traversal (cache Tree, Context, Options) Re-export cache::{Tree, traverse::{Context, Options}} at cache so external ca
<csr-id-40a4ca7d1802361850a6a4a6aef91c8c2c5d730f/> expose index-less parallel pack traversal (cache Tree, Context, Options)
Re-export cache::delta::{Tree, traverse::{Context, Options}} at cache so
external callers can build a delta Tree directly from a pack header scan and
resolve all of its objects in parallel via Tree::traverse -- without a
pre-built .idx. This is the index-less companion to the already-public,
idx-verified index::File::traverse_with_index.
Minimal surface by design: the delta module stays pub(crate), so only the
three already-documented items a caller names become public -- pub mod delta
would instead expose the module's internal error types and constructors and fail
#![deny(missing_docs)]. Tree::from_offsets_in_pack (already public) builds the
tree straight from the pack; with_capacity/add_root/add_child remain for
callers who assemble it by ascending pack offset themselves.
Additive and non-breaking; cargo doc is clean under -D warnings, and a
doc-test anchors the index-less path (Tree::from_offsets_in_pack). Shape
approved in #2922.
<csr-id-916f5d8c32deb0a03bb6a01c8d8e9c4b6c69d186/> validate delta base sizes during entry decoding <!-- agent --> Reject delta instruction streams whose declared base size differs from the resolved base object. The regression demonstrates that a REF_DELTA could previously reconstruct successfully despite lying about its base size.
Git reference: patch_delta() compares the decoded source size with src_size before applying instructions, observed at cf5497b14c5a.
<csr-id-d7bbb8e053c3ce9c24fe0a4bf1aaaca7b7a58270/> resolve delta trees lazily in a lock-free pool <!-- agent --> --- Summary of findings
The lazy, stealable resolver fixes the pathological phpstan pack without regressing the Linux best case. At 16 threads, phpstan falls from the 98.38-second baseline resolver time to 22.70 seconds, while the Linux best-case resolver improves slightly from 11.81 to 11.33 seconds. Charged peak memory on phpstan falls from 13.14 GB to 2.16 GB, and the Linux best case falls from 1.96 GB to 1.76 GB.
Across the measured packs, gix resolver speedup at 16 threads ranges from 10.80x to 12.48x over its new serial path. On phpstan, gix is 1.07x faster than Git at one thread and 2.31x faster at 16 threads. On the identical 10.9M-object Linux payload, gix's SHA-256 resolver is 2.72x to 2.99x faster than its SHA-1 resolver, confirming that the slow SHA-1 implementation dominates much of the remaining absolute cost.
The Git SHA-256 comparison was initially unfair: the local Git build used the portable SHA256_BLK backend while RustCrypto selected ARMv8 SHA-2 instructions. Rebuilding Git with OpenSSL reduced its eight-thread wall time from 144.15 to 40.45 seconds and user CPU time from 471.94 to 99.81 seconds. All final Git SHA-256 comparisons use that faster backend.
--- Pathological phpstan pack
The pack contains 100.7k objects but expands to 174.6 GB because one root ends in a deep, expensive delta tree. The previous resolver takes 98.53 seconds wall clock and has a 13.14 GB peak memory footprint.
The new resolver takes 283.41, 71.94, 36.42, and 22.83 seconds wall clock at 1, 4, 8, and 16 threads. Its resolver speedups are 3.94x, 7.81x, and 12.48x relative to one thread, remaining nearly linear through eight threads. Peak memory footprint is 0.68, 0.75, 1.17, and 2.16 GB respectively. Thus the fastest run is 4.31x faster than the old wall time while using 84% less charged peak memory.
Git was measured at 1, 2, 4, 8, and 16 threads while keeping its aggregate delta-base cache allowance at least as large as the default eight-thread allowance of 8 times 96 MiB. Git takes 304.51, 168.06, 98.80, 65.41, and 52.79 seconds, for 5.77x one-to-sixteen-thread scaling. At equal thread counts, gix is 1.07x, 1.37x, 1.80x, and 2.31x faster at 1, 4, 8, and 16 threads.
At the nearest wall-time operating points, gix at four threads takes 71.94 seconds with a 0.75 GB peak footprint, while Git at eight threads takes 65.41 seconds with a 1.90 GB peak footprint. gix therefore uses 61% less charged working memory near Git-matching throughput.
--- Linux best-case pack
The 7.6M-object fixture expands to 95.6 GB and guards the already-friendly case. The old implementation takes 13.68 seconds wall clock and 11.81 seconds in the resolver. The new implementation takes 137.48, 36.75, 19.81, and 13.16 seconds wall clock at 1, 4, 8, and 16 threads; resolver scaling reaches 11.97x at 16 threads.
The 16-thread result is slightly faster than the old implementation, with 11.33 seconds in the resolver, and lowers peak footprint from 1.96 to 1.76 GB. Peak footprint remains effectively constant from one through 16 threads. Git takes 156.65, 60.48, 45.48, and 42.97 seconds at the same thread counts, so gix's advantage grows from 1.14x to 3.27x.
--- SHA-1 and SHA-256 sibling packs
The sibling packs contain 10.9M objects, expand to 146.7 GB, and preserve the same compressed payload and delta topology. This isolates hashing from work distribution.
For SHA-1, gix takes 212.57, 56.58, 30.60, and 20.74 seconds wall clock at 1, 4, 8, and 16 threads. Resolver scaling reaches 11.36x at 16 threads. The default eight-thread Git run takes 125.11 seconds, while gix takes 30.60 seconds at eight threads and 20.74 seconds at 16.
For SHA-256, gix takes 72.47, 21.83, 12.58, and 8.67 seconds wall clock. Resolver scaling reaches 10.80x at 16 threads. Against OpenSSL-backed Git, gix is 1.21x, 2.02x, 3.22x, and 5.61x faster at equal thread counts. Git improves from 87.85 seconds at one thread to 40.45 seconds at eight, then regresses to 48.68 seconds at 16 with 204.81 seconds of system CPU.
--- Hasher and scheduler interpretation
On identical pack data, gix SHA-256 resolver time is 70.31 seconds at one thread versus 210.35 seconds for SHA-1, and 6.51 seconds at 16 threads versus 18.51 seconds for SHA-1. SHA-256 is therefore 2.99x faster serially and 2.84x faster at 16 threads, with a 2.72x to 2.99x advantage throughout the measured range.
The similar 10.80x to 12.48x gix scaling across pathological, best-case, SHA-1, and SHA-256 packs argues against lock contention being the main high-thread limitation. The remaining flattening is consistent with finite parallel work, scheduling overhead, memory bandwidth, and hashing cost.
--- Memory interpretation
The resolver no longer materializes the unresolved internal-node frontier. It resolves a child only when a worker starts it, shares immutable bases between sibling tasks, and recycles the final base reference. A linear chain therefore needs roughly two object buffers per active worker instead of retaining every intermediate base.
macOS max RSS includes clean file-backed pages. gix maps the pack, so its reported RSS can include most of a multi-gigabyte pack even when those pages are reclaimable; Git reads through bounded pread buffers. Peak memory footprint better represents charged working memory here. On phpstan, gix max RSS grows from 7.20 to 8.90 GB across 1 to 16 threads, but charged footprint grows from only 0.68 to 2.16 GB. On the Linux best case, footprint stays at 1.76 GB across the same range.
--- Implementation
Use lock-free local deques and a shared root injector so idle workers can steal branches from the last expensive root without every worker hoarding roots. Enable this resolver through the gix parallel feature.
Regression tests prove that internal siblings are not all materialized before descent and that two workers can concurrently resolve children from one remaining root. The scheduler follows the proven Git index-pack principles of bounded live delta bases and keeping independent delta work available, informed by builtin/index-pack.c at cf5497b14c.
<csr-id-64b9efefb9a538a0c40f3413d6c758a17bbc2daf/> resolve in-pack ref deltas while indexing Valid packs may encode a base inside the same pack with REF_DELTA and may place the delta before that base. The streaming lookup previously treated an object-database miss as fatal, while index construction rejected every remaining ref delta.
Keep lookup misses in the stream and park ref-delta children by base object ID until traversal resolves the matching object. This preserves external thin-pack injection and reuses the existing delta tree for forward and chained references.
Git reference: a23bace963d508bd96983cc637131392d3face18, builtin/index-pack.c and t/t5300-pack-object.sh.
object_hash from Options
This forces callers to explicitly choose a hash, eliminating the risk
that Sha1 is implicitly chosen for them in SHA-1/SHA-256 builds
through Default::default() without them being aware of it or noticing.
This comes at the cost of callers having to always pass the hash, but
that seems like a reasonable trade-off.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
64b9efe)d7bbb8e)916f5d8)ebe9095)b2d919a)object_hash from Options (c055803)f3bbfad)gix-testtools (0cbe539)c4426f0)dcb8ccd)40a4ca7)dd8c759)gix-odb (1dc741f)9c12c2d)cc3ee80)ab4fcb0)4a6cf9d)9b787f6)ae8845a)
</details>cap remaining gix-pack allocations
gix_hash::Kind in
Entry::from_bytes() instead of accepting a raw hash length. This keeps callers
from passing unsupported or inconsistent hash lengths and lets the decoder derive
the ref-delta base id length from the selected object format.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
0ba5af4)5850141)ace687c)842bc44)cb6ec7d)f0ec710)c16b5a1)43ff87a)cf3053a)27aec47)6e1c4a2)f1ac335)f26c18a)6d95da6)gix-features, use gix-zlib accordingly. (9c2977a)0052e42)09642c6)223f673)6391b3e)938506b)5d4c52b)b69f0a6)b2a0fd6)3d511c3)d785bbc)mmap2 to avoid potential for unsoundness (eda41b1)1b1541e)
</details>cap aggregate delta data allocation in gix-pack A ClusterFuzz data_file testcase could build a malformed delta chain whose individual entry sizes stay
<csr-id-6de909bec54270cf2433eb944af20ed0e7bd7467/> cap aggregate delta data allocation in gix-pack A ClusterFuzz data_file testcase could build a malformed delta chain whose individual entry sizes stayed below the configured fuzz allocation cap, but whose aggregate decompressed delta payload size reached multi-gigabyte scale. The fuzz harness then attempted to reserve that aggregate buffer and aborted with libFuzzer out-of-memory.
Reject aggregate delta payload sizes once they exceed File::with_alloc_limit_bytes(), matching the existing protection for individual decoded object sizes. Add the minimized ClusterFuzz testcase to the data_file artefacts so the known input remains available to the fuzz target and artifact smoke test.
index::Version::hash() method.
It's not useful either as there is no relationship between the Version
of the index file and the hash to use.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
6428edc)cdafa6a)14025af)bbf6fe3)aa319aa)70d38bf)6de909b)4f862a5)index::Version::hash() method. (ee91e31)gix-pack (4f1bb83)f7d4f33)
</details>Add SHA-256 support This means that if compiled in, SHA-256 packs can now be read and used.
<csr-id-3e389020600415ad9fd0de38835534bfe4889ba0/> reject truncated delta headers within their own data range ClusterFuzz reported a panic in gix-pack's data_file fuzz target for testcase clusterfuzz-testcase-minimized-gix-pack-data_file-6499869547364352. The minimized pack could reach File::decode_entry() and panic while applying a relocated delta instruction range with the slice error 'slice index starts at 11 but ends at 9'.
The first delta decoding pass stored all decompressed delta instructions contiguously, but decoded each delta's base and result size headers from the remaining combined instruction buffer. If a malformed delta header continued past that delta's declared decompressed size, header parsing could consume bytes from the following delta and produce an invalid instruction range for the current delta.
Limit delta size-header parsing to the current delta's decompressed instruction range so truncated malformed headers become delta corruption errors instead of cross-delta ranges. Add the ClusterFuzz artifact and a regression that exercises the same offsets as the data_file fuzz target.
It's notable that it doesn't reproduce when run in debug mode or with the test-suite, but it does reproduce with the fuzzer, and it's fixed now (hence does not reproduce). To keep the suite simple, I removed the extra test.
<csr-id-176d07b9bd00bc83d310f0034452a1448414c55f/> accept non-canonical pack entry size headers C Git accepts overlong pack entry size encodings, and real servers can send them when reusing existing pack data. Accept these headers while recording the actual header length consumed from the pack.
Keeping the actual header length avoids recomputing a canonical length from the decoded size, which would break pack offset reconstruction and ofs-delta base offset calculations for non-canonical entries.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
10c58bb)176d07b)575113d)gix-testtools (ce9e6bd)8af2691)04c894e)278d7ec)0ec3bb7)justfile and automate feature tests (db7b97b)24cdbb0)c5db356)af84752)89a27bf)8e562c7)3e38902)ab2fee1)2cb17b2)e10d5f6)3675a8d)23af41a)gix_object::Data (4309fa4)gix-testtoolsand rename hash_kind -> object_hash (d9648e8)adb8328)
</details>3 commits contributed to the release over the course of 2 calendar days.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
53f880c)gix-object (91bfab0)4d5ba23)
</details>enforce the specification of alloc_init_bytes to handle untrusted input This breaking change is intended to force a decision about how much memory all…
alloc_init_bytes to handle untrusted input
This breaking change is intended to force a decision about how
much memory allocation an untrusted party can command by tempering
with binary file formats.data
Previously we'd hardcode a memory map, but now this is parameterised
with a trait to allow accessing data in a more abstract form.
This is primarily meant to speedup fuzz testing,
but may also be useful later for portability.leb64 and secure leb64_from_read from overflows<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
f9fbcba)63b8419)7429b15)e64e3b8)0396152)alloc_init_bytes to handle untrusted input (91d0c26)1dc8ede)leb64 and secure leb64_from_read from overflows (c0ab0f4)cf5279d)data (3ef3916)392336f)7d50c30)d9b358b)gix-object (6df1d55)2a5db88)e4f380e)444a92b)3e05ca3)package.include patterns more specific so they don't match ignored files (c2c917f)98bae84)
</details>add sha1 and sha256 features to gix. This way one can control which hashes are compiled in exactly, while having reasonable defaults automatically.
sha1 and sha256 features to gix.
This way one can control which hashes are compiled in exactly,
while having reasonable defaults automatically.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
c389a2c)6183fd0)6bdb331)e8bf096)sha1 and sha256 features to gix. (3832916)gix-hash (e71c703)f7277f3)
</details>skip subtrees with identical oids Add a short-circuit check to the core gix-diff algorithm to skip identical subtrees based on oid. This substantially
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
ecf90fc)e441fa9)e63d487)228caf7)5932e88)GIX_TEST_FIXTURE_HASH for gix-commitgraph and gix-pack. (d51b858)9327b73)
</details>6 commits contributed to the release over the course of 19 calendar days.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
d66ac10)8bceefb)7ce3c55)e8612b5)gix-error (a304f13)f7d0975)
</details>Use std::ControlFlow where possible
std::ops::ControlFlow where possiblegix-error instead of thiserror<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
cbbdef5)562e684)std::ops::ControlFlow where possible (5c1bd03)4c6a7a7)gix-chunk (e6e90ff)gix-error instead of thiserror (0c68131)0a333e5)c663b3f)
</details>2 commits contributed to the release.
<csr-id-d99352bd3b7c75791826b40df20afdd4b0d8b26e/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
7be8f90)c0cf38f)d99352b)211b4fb)
</details>retry the current commit when graph parents are corrupt
retry the current commit when graph parents are corrupt
Both Simple traversal paths pop the current commit before decoding its
cached parents. If parent iteration fails, dropping the cache and recursively
requesting the next commit skips the popped commit and its entire ancestry.
With an out-of-range inline parent, a two-commit gix revision walk could
therefore succeed with an empty result.
Put the current commit back at the front of the topology queue or into the
date queue with its existing priority, then continue the loop after disabling
the cache. The object database retries that same commit without recursion;
any object lookup or decoding failure is still propagated.
The regression corrupts a disposable graph, confirms that Git rejects it,
and checks that the existing gix fallback policy preserves both commits
across breadth-first traversal, both date orders, cutoff sorting, and
first-parent traversal.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
115e208)d6c5b9d)
</details>5 commits contributed to the release over the course of 13 calendar days.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
21fecdf)
</details>allow configuring the reqwest HTTP client
allow configuring the reqwest HTTP client
Applications need to supply a custom TLS configuration, including an explicit
cryptographic provider, before reqwest builds its HTTP client. The existing
request callback runs too late to configure that client.
Add reqwest::Options::configure_client and build the client lazily when the
first request makes the HTTP options available. Reuse that client for later
requests, retaining the configured options if a failed worker is restarted.
Apply Git's redirect policy after customization so the callback cannot replace
its safeguards for configured request headers or unsafe redirects.
Adding this public option requires existing complete Options literals to
specify configure_client or use ..Default::default().
reqwest HTTP client (7111e30)Keep rust workspace tests inside disposable repositories and isolated environments
Direct Git launches inherited repository selectors and user configuration even
when tests supplied a fixture working directory. Tests of default-environment
APIs and local Git transports also shared the runner's environment. A few
journey tests wrote beneath source directories or used the source checkout as
the repository under test.
Use the shared gix-testtools Git command builder for subprocess setup, isolated
repository options for fixtures, and isolated child processes where the real
environment-reading API must be exercised. Scope CWD changes, copy the fixture
used by an object-write test, and run shell journeys through jtt run. Keep
journey worktrees and example output within their disposable sandboxes and
replace the attributes checkout test with a representative fixture repository.
Prompt examples also run in isolated children and must build successfully; the
old tests could ignore build failures and execute stale cached binaries.
The affected Rust crate suites, internal test-tool build, and max-pure journey
suite pass from a source copy without Git metadata. Signing and Git-daemon
checks use only disposable keys, repositories, and local sockets.
migrate errors to gix-error
rubberstamp
raise MSRV to Rust 1.88
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
82ff92f)fbf9c39)
</details>3 commits contributed to the release over the course of 1 calendar day.
<csr-id-6f469a6fea59c88e6c69a5f94b0bc8a5977cb75b/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
bdcce5f)doc_auto_cfg feature to fix docs.rs documentation. (6f469a6)3313233)
</details>Headline benefits of this extra-curricular release are various security improvements, and greatly improved error handling and consisstency. Plumbing c
Headline benefits of this extra-curricular release are various security improvements,
and greatly improved error handling and consisstency. Plumbing crates for instance will now
use a standard std::error::Error compatible error type.
Besides that, the tracing of all commands was greatly improved.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
49f8d53)89fb308)4da2927)b188a7d)6cc8464)9365cc3)838ff95)752d6dc)0e7aa81)dc7343c)79ba9d0)202bc6d)d64f257)473fe52)428412c)784c046)
</details>Headline benefits of this extra-curricular release are various security improvements, and greatly improved error handling and consisstency. Plumbing c
Headline benefits of this extra-curricular release are various security improvements,
and greatly improved error handling and consisstency. Plumbing crates for instance will now
use a standard std::error::Error compatible error type.
Besides that, the tracing of all commands was greatly improved.
remove full-screen progress mode <!-- agent --> Full-screen progress required separate command and UI threads plus channel coordination, while line progress already covers long-running commands and works with independently buffered tracing.
Remove --progress, --progress-keep-open, and the prodash-render-tui feature. Keep synchronous line progress as the only progress UI.
make trace output independently configurable <!-- Byron -->
Quick review only, as it's mostly about tracing and gix which both aren't super
important, and are seemingly working well. An extra agent-review couldn't
find anything either.
<!-- agent -->
Replace the old boolean trace switch with a bounded -t/--trace counter:
forest INFO and DEBUG at one or two occurrences, then flat DEBUG and TRACE at
three or four. Keep --threads long-only and route tracing independently of
verbose and progress rendering, including commands with bespoke execution paths
and corpus runs.
Initialize display tracing once before repository discovery, buffer it outside
prodash, and emit it after command and progress teardown. Format forest and
flat output with ANSI, then adapt the final stderr write so terminals get color
while pipes and NO_COLOR remain plain.
Corpus worker subscribers share that sink while retaining their complete
JSON trace storage, and flat modes include completed spans. gix tix remains
delegated to Tix so it installs only one subscriber.
add and remove linked worktrees from the CLI <!-- Byron -->
rubber-stamp mostly for gitoxide-core changes, did look at it though just to be sure.
Journey test was present but removed.
<!-- agent -->
The CLI previously exposed only worktree listing, despite gix supporting
both checkout and safe removal. Add worktree add (alias create) and
worktree remove using those APIs.
Creation follows Git's local-branch and detached-revision selection, including
destination-basename defaults and -b. It retains new branches after setup
failures, uses the existing listing format on success, and reports incomplete
checkouts. Removal maps repeated -f flags to dirty-worktree and lock overrides
while retaining branches.
Both commands use the existing repository, configuration, and progress options. Unsupported JSON output is rejected before mutation.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
AsHashKind (106db87)gix free index (a8e3143)<csr-id-fce70950006892f51b32af233656be6fe5de9df3/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
5a919c4)65037b5)dab97f7)a9a8ea1)c2eb0c1)bfb1c34)52906fb)56ca4bf)fce7095)f8d7c0a)151e3a5)c3f06ae)d2b4c44)8d4c4d1)
</details>3 commits contributed to the release over the course of 1 calendar day.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
e104545)5f7f805)631f07a)
</details>Breaking change: gix_worktree::prepare() now accepts the destination first, followed by an add::Context containing the common Git directory and an opt…
preserve sparse-checkout state when adding worktrees
Looked at everything carefully, but was despite test refactoring, I went
through those more quickly. The benchmark I only skimmed.
Adding a linked worktree from a sparse checkout copied its configuration
without its patterns, then materialized the entire target tree. Copy active
info/sparse-checkout patterns from the source worktree's private Git directory
during preparation, including when checkout is deferred. Missing patterns are
skipped, empty files are retained, and copy failures roll back registration.
Expose pattern loading and full-index selection through gix_worktree::sparse.
Initial checkout in gix reads the destination's current patterns and marks
excluded target-tree entries with SKIP_WORKTREE and EXTENDED before writing
files and persisting the index. Support cone and non-cone patterns, case
folding, invalid-cone fallback, root entries, and directory pruning. Sparse
index generation and updates to existing checkouts remain outside this API.
Reuse matching decisions for shared directory ancestors with storage bounded
by path depth, and skip index traversal for a validated full-cone selection.
Cover pattern matching with 18 Git-generated baseline scenarios, regression
tests for cache boundaries and ordering, and integration tests for main and
linked sources, deferred checkout, missing or disabled patterns, and rollback.
Add Criterion benchmarks for cone and non-cone selection on synthetic indices
with 20,000 to 200,000 entries, checking membership and flags before timing.
Breaking change: gix_worktree::add::prepare() now accepts the destination
first, followed by an add::Context containing the common Git directory and
an optional source private Git directory, then its existing options. Use
add::Context::new(common_dir) when pattern inheritance is not wanted.
make bounded removal retries configurable
Went through the tests pretty quickly, as is increasingly common.
Expose Options::retry_delay alongside max_retries. The default is 10 ms;
Duration::ZERO retries immediately, and callers can choose a longer delay for
longer (expected) write bursts. Only retryable DirectoryNotEmpty errors with
attempts remaining wait, before restarting traversal. Successful first-pass
removal, zero-retry operation, and terminal failures incur no pause. The default
two retries request at most 20 ms of sleep per root; filesystem operations and
scheduler delays have no wall-clock bound.
Remove the shrinking-entry-count guard. A writer can replace every deleted file
with a new one, leaving the scan count unchanged even though deletion succeeds
for every scanned entry. The old guard then stops before the configured retry
allowance is used. A deterministic regression writes one replacement after each
of the first two scans: the old loop fails on the second pass; the corrected
loop succeeds on the third, even with a zero delay. The retry count bounds
shrinking, equal, and growing workloads without trying to infer writer progress.
Deletion duration or entry counts cannot tell when an independent writer will
finish. Keep the delay an explicit caller policy instead of introducing another
heuristic tied to tree size. No finite retry policy guarantees completion for
arbitrary finite or indefinite writers. Existing porcelain Target::options()
and the re-exported Options expose this setting without another configuration
layer. Downstream exhaustive Options literals must add retry_delay or use
..Default::default().
cap automatic macOS removal at four workers
This is an APFS problem, which is also noticeable when scanning through
metadata, which now caps at 8 at least in dua.
On Linux Ext4 no such limits exist for all I know, so this commit is
a macOS specific fix to ensure decent performance.
Resolve automatic removal concurrency to at most four available logical CPUs
on macOS. None and Some(0) stay automatic; positive explicit limits remain
unchanged, as do defaults on other platforms. Apply the cap in the shared
plumbing options so repository, proxy, and CLI removal inherit it.
Sixteen concurrent unlink workers contend on native APFS. A dedicated comparison
of three and four workers favors four in all 12 matched pairs, including trees
created by Git, gix, and APFS cloning. Median wall seconds (range), four samples
per cell:
| Creator | 3 workers | 4 workers | Less elapsed time |
|---|---|---|---|
| Git | 2.013 (1.878–2.059) | 1.842 (1.820–1.856) | 8.49% |
| gix | 2.873 (2.849–2.890) | 2.597 (2.577–2.673) | 9.59% |
| APFS CoW | 1.577 (1.571–1.594) | 1.429 (1.420–1.432) | 9.42% |
Median per-round 3/4 time ratios are 1.087×, 1.101×, and 1.107× respectively.
One additional single-worker control per creator took 3.258, 3.911, and 2.781 s:
the four-worker medians are 1.77×, 1.51×, and 1.95× faster. Those controls ran
last and are diagnostic, not paired repetitions.
The preceding native worker sweep measured sixteen-worker medians of 5.018 s
for Git-created and 5.819 s for gix-created trees (two samples each). These are
separate measurements, not additional pairs in this comparison. GitButler's
image-derived one-worker override and the old all-core default both give up
native-volume throughput.
2026-10-09, Apple M4 Max, 16 logical CPUs, 64 GiB RAM, macOS 27.0.1 (26A434),
native APFS under /private/tmp; Rust 1.96.1 release probe and Apple Git 2.54.0.
The cached probe uses local Gitoxide 00bb1af, with max-performance, parallel,
both pack-cache variants, SHA-1, and worktree mutation. Explicit removal thread
limits exercise the same deletion path as the new automatic cap. The retry
pause only applies after an error; no concurrent writes were injected here.
Borrow objects read-only from
/Users/byron/dev/git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux, commit
707df3375124b51048233625a7e1c801e8c8a7fd, using independent seed metadata.
The clean seed checks out tree 176a057fce1932fe5057196e63c2a636e1a3cdab:
88,821 tracked files / 1,509,611,994 blob bytes. Native case-insensitive APFS
requires omitting thirteen later case-fold collisions in git ls-tree -r
order: 15,749 bytes, 0.00104% of the source tree.
Each of four rounds creates fresh Git, gix, and CoW destinations with both
removal counts; odd rounds reverse all case ordering. Git checkout uses one
worker, gix its default sixteen. CoW registers with Git without checkout,
uses clonefile() for regular files and index, and recreates symlinks. This is
a deletion fixture, not a timing of GitButler's entire CoW creation workflow.
A small byte/symlink self-check validates the fixture copier.
Synchronize with sync_volume_np(SYNC_VOLUME_WAIT) before creation and removal,
outside timers. Validate clean status before forced removal, excluding status
cost from the removal timer. All 27 samples removed both checkout and private
administration. No build or other benchmark overlapped measurements. One
0.24 s read-side inventory overlapped the second-round CoW/3 case; excluding
that pair leaves medians 1.577 vs 1.430 s, with four winning all remaining pairs.
These are warm-cache timings on one native APFS machine. No new disk-space,
cold-cache, durable-write, or other-platform measurement was made. The old
sparse-image results are unsuitable for choosing native macOS concurrency.
The updated automatic-resolution test failed with 16 instead of 4 before the
change. Removal and bounded late-write regressions pass, including explicit
limits above the cap. Keep results here; discard temporary fixtures and logs.
classify sparse gitlinks as directories on every host
Non-cone directory-only patterns should treat gitlinks as directories on
every host, including when the destination is initially empty. The original
matcher instead hardcoded them as non-directories, reproducing the result
of Git's little-endian ce_to_dtype() bug.
Expose gix_index::entry::Mode::is_dir_for_git_matching() using decoded
modes directly, and document why Git's extra ntohl() conversion is wrong.
Use this classification for sparse matching, so /*/ includes an absent
gitlink and !/module/ excludes it consistently across hosts.
Keep Git-generated baselines while working around its conversion bug:
pre-create the empty gitlink directory only for the reference operation,
allowing Git's filesystem fallback to recover its intended classification.
The gix index-only regression and initial checkout still start without it.
Check included gitlinks as directories rather than reading them as files.
raise MSRV to Rust 1.88
<!-- agent -->
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
gix editor command
<!-- agent -->
Provide a small plumbing command that opens zero or more paths with
Repository::editor()'s Git-compatible selection. Prepare editor definitions with
command-aware argument splitting so PATH-resolved programs run directly while
actual shell syntax still uses the shell, reject explicitly empty settings, and
honor ':' as the no-op editor.<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix editor command (d930653)
</details>A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
cc5b696)49fa9f3)db0b095)0bf84db)3b1bef7)c3c6504)5823b22)Cargo.toml files in workspace to gix-features bump (6315536)b5e9059)68e6b2e)1612c73)fc5faf2)420e730)
</details>remove gix tix screen selection Remove the plumbing CLI screen-mode option now that the interactive history always owns the alternate screen. The auto
remove gix tix screen selection Remove the plumbing CLI screen-mode option now that the interactive history always owns the alternate screen. The automatic and half-screen modes depended on inline rendering behavior that is being retired from gix-tix.
Stop accepting --screen for gix tix and its aliases, and remove the mode validation and translation into gix-tix. Construct the reduced gix-tix options directly from the remaining quit and hidden-revision arguments. Keep a command-line regression assertion so the removed option cannot silently return as an ignored argument.
delegate tix commands through gix-tix Embed the complete gix-tix Clap platform in the plumbing command instead of maintaining a second parser and dispatcher. This keeps standalone and plumbing invocations aligned, adds worktree-tip parity, and makes amend and spill available through gix tix without duplicating their execution logic.
show other worktree checkouts in tix Decorate commits checked out by other main or linked worktrees with light-blue name@ labels, replacing the ordinary branch label. Keep the checkout from which tix was opened represented by the graph @ marker and its ordinary local reference, and use directory basenames for other detached worktrees. Keep other-worktree labels visible on the selected row when references are hidden.
Add -w/--worktrees to include every successfully resolved worktree HEAD, including the current checkout, alongside implicit or explicit traversal tips without weakening hidden-revision exclusions. Discover worktrees from their private Git metadata so stale checkout directories remain useful, while malformed, unborn, or inaccessible entries are logged and skipped.
Watch linked HEAD and worktree membership changes so decorations and optional tips stay current, while ignoring unrelated linked indexes, logs, and metadata. Document the behavior and cover current, attached, detached, stale, malformed, hidden-reference, CLI, and watcher cases.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
remove linked worktrees safely
Provide the gix-worktree plumbing needed to recursively remove a linked
checkout and its private Git directory without following symbolic links or
crossing Unix mount boundaries.
Use dua-core for traversal and parallel deletion with progress reporting,
bounded retries for directory-not-empty races, and errors distinguishing
checkout and administration failures. Keep repository validation, dirty-state
checks, lock policy, and branch decisions with higher-level callers.
create linked worktrees
Add transactional linked-worktree administration with rollback, then compose
it in gix with existing reference, checkout, index, progress, and interruption
machinery.
Support attached and detached worktrees from bare or non-bare repositories while
rejecting occupied branches and destinations. Keep move, remove, repair, hooks,
sparse checkout, and relative-link configuration out of scope.
skip symlinked ignore files and protect checkout writes
A tracked .gitignore symlink was absent from the skip-worktree ID
mapping, so ordinary ignore lookup could parse arbitrary external files.
An endless target could also exhaust memory.
Always request no-follow reads for worktree .gitignore files, regardless
of their index entry. Use gix_fs::open_read_only_no_follow() when pattern
loading disallows symlinks, treating skipped links as absent files so
skip-worktree blob fallback remains available. Discard partially read
pattern bytes on failure.
Use the same helper for the worktree .mailmap, reporting symlinks as
errors without reading their targets and identifying worktree read errors
correctly. Explicitly configured pattern files and mailmap.file may
still follow symlinks.
On Windows, inspect nonexclusive checkout handles for symlinks before
truncating with set_len(0) or writing. Retain forced unlink-and-replace
behavior and route symlinks checked out as ordinary files through the
same validated open path. No-follow protection applies to the final path
component on Unix and Windows; leading components can still be symlinks.
Compare ignore lookup and skip-worktree fallback against isolated Git
check-ignore calls. Cover mailmap policies and checkout collisions with
file, directory, and dangling symlink targets, including links created
after the Windows path check and validation before truncation.
git_hash::verify::Error instead of bespoke duplicated versions
thereof, and occasionally returning an ObjectId instead of ()
for convenience.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
d248e3d)ada5a94)b41312b)38dff41)4660f7a)4501086)gix_object::{try_ =>}compute_hash users (3d7e379)fbf6cc8)5095f44)4f2b649)54e5764)6c02b0a)gix_hash::Hasher::finalize() (4e935ce)gix_hash (baa1430)7b17da6), just before ) on same line in function calls (66a5ae1)16a248b)8e96ed3)8df0db2)
</details>support cloning a single revision <!-- agent --> A full object ID passed through with_ref_name() produced an object-ID refspec mapping and panicked wh
support cloning a single revision <!-- agent --> A full object ID passed through with_ref_name() produced an object-ID refspec mapping and panicked while clone assumed every mapping had a name. Branch and tag checkout also retained ordinary clone tracking semantics instead of offering a single-revision mode.
Add PrepareFetch::with_revision() and gix clone --revision for full refs, HEAD, and full object IDs. Revision clones use a one source-only implicit refspec, detach HEAD to the fetched commit, create no ordinary refs, persist no fetch refspec, and disable tag following. Existing with_ref_name() and --ref behavior stays unchanged.
This follows Git commit 337855629f59 (builtin/clone: teach git-clone(1) the --revision= option) and its t/t5621-clone-revision.sh behavior.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix-date (613ff86)keep descendants of ignored directories excluded
I looked at this carefully, and ended up having a better comment
to provide an example for what this covers.
Ultimately, this is test extension is still minimal compared to the
behaviour that it enables, scoped status, but I think it's fine
in there interest of less code.
Scoped status walks can enter an ignored directory when a pathspec
matches. If a deeper directory matches a negated rule such as !out/,
the ignore stack previously stopped at that nearer match and exposed its
contents as untracked. Tix uses scoped walks for filesystem refreshes,
so ignored Cargo output appeared after unrelated worktree events and
disappeared after a full refresh.
Continue through negated directory matches to check excluded ancestors.
Keep the nearest negation as the fallback when no ancestor excludes the
path, and preserve the precedence of positive directory matches,
including precious-file handling.
Extend the Git-reference fixture below a negated child directory and
refresh both hash archives. The extended fixture fails with the previous
matcher. All directory-walk, status, and worktree tests pass, as do the
SHA-256 ignore baseline, formatting, and Clippy with -D warnings.
Clippy retains the workspace's existing removed-lint warning.
raise MSRV to Rust 1.88
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
<csr-id-17835bccb066bbc47cc137e8ec5d9fe7d5665af0/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
dea106a)1f6390c)7ec21bb)rust-version to 1.70 (17835bc)1ca480a)9193b05)0ab4f64)gix-traverse (1de4e70)d22937f)
</details>add configurable terminal screen modes.
add configurable terminal screen modes.
hide revision ancestry in tix
add tix to the gix CLI
add explicit gix config show subcommand.
add gix config list for showing the git-configuration files.
add the gix config fmt subcommand
Expose the gix-config whitespace formatter as
gix config fmt [--in-place] [in-file] [out-file]: with no in-file it
formats the repository-local configuration, and with no out-file it writes
to stdout. The repository is only opened when the repository-local
configuration is needed, so formatting an explicit file works outside a
repository too.
add gix status --untracked flag
This allows to control how untracked files are folded, but also
can completely turn dirwalking off.
add gix free remote refs to list and write remote refs.
Add gix free remote refs to perform an upload-pack handshake, discover
the remote reference advertisement, and print the refs without negotiating
or receiving a pack.
The command can also write the advertised refs into a standalone ref store:
gix free remote refs --refs-directory out-refs <url>
preserve multiple remote URLs
Configured remotes can have multiple remote.<name>.url or remote.<name>.pushUrl
values, but gix::Remote kept only one value because lookup used the singular
config accessor. That meant the most recent config value won, while Git exposes
every effective URL and uses the first one for the singular get-url form.
Add ordered URL storage to gix::Remote and expose Remote::urls(Direction) for
all effective URLs. Keep Remote::url(Direction) as the singular compatibility
API, now returning the first effective URL. Fetch URLs used as push fallbacks
try pushInsteadOf first and then insteadOf, matching Git behavior.
The regression fixture records Git 2.50.1 behavior with git remote get-url: without --all it prints the first configured URL, and with --all it prints all configured URLs in order for both fetch and push.
make exclude queries index-aware <!-- agent -->
gix exclude query reported ignore matches for tracked files and directories
containing tracked entries, unlike git check-ignore. Positional arguments were
also interpreted as pathspecs while stdin supplied paths.
Treat positional and stdin input uniformly as paths, normalize them relative to the repository, and suppress ignore matches for indexed files or directories containing indexed entries. This changes positional arguments from pathspecs to paths.
Add journey coverage for tracked and untracked paths below ignored directories, stdin from a nested working directory, ignore-pattern display, and positional output order.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix config list for showing the git-configuration files. (bd595de)gix config fmt subcommand (87a5e89)gix-attributes (e11d7a2)gix-config (376e946)gix status --untracked flag (9316ff8)gix free remote refs to list and write remote refs. (6767b30)
</details>gix-testtools (0cbe539)Store::at() where possible (17fea2a)GIX_TEST_FIXTURE_HASH for more tests (bbea2c4)A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
beb0ea8)c1ba571)7ea8582)e8b3b41)
</details>Strange - automatic release notes weren't generated or didn't arrive here. Have to look into that. Meantime, let's use the GitHub generated notes.
Strange - automatic release notes weren't generated or didn't arrive here. Have to look into that. Meantime, let's use the GitHub generated notes.
fuzzed_timeout on windows-11-arm by @EliahKagan in https://github.com/GitoxideLabs/gitoxide/pull/2619gix-pack by @cruessler in https://github.com/GitoxideLabs/gitoxide/pull/2602GIX_TEST_FIXTURE_HASH for gix-index tests by @cruessler in https://github.com/GitoxideLabs/gitoxide/pull/2598should_interrupt per entry in index_as_worktree by @sharifhsn in https://github.com/GitoxideLabs/gitoxide/pull/2631Full Changelog: https://github.com/GitoxideLabs/gitoxide/compare/v0.54.0...v0.55.0
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
4000197)8ce4912)bc9d994)54ea266)664e28c)aeaebec)dc0a73a)<= on 32-bit for some size assertions (fc13fc3)7a40648)0f0e4fe)1411289)izip! doctest ignore (0783e9c)db5c9cf)
</details>gix dirwalk as a way to run gix-dir directly This is mostly for testing walks specifically, without needing them as part of gix clean or gix status.
gix dirwalk as a way to run gix-dir directly
This is mostly for testing walks specifically, without needing them as part
of gix clean or gix status.gix merge tree --message '' to allow creating commits
This is useful for controlling cherry-picks precisely.<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix dirwalk as a way to run gix-dir directly (add4cf4)gix merge tree --message '' to allow creating commits (6ab327e)
</details><csr-id-64ff0a77062d35add1a2dd422bb61075647d1a36/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 2 times to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
f1364dc)3f7e8ee)b36d7ef)2d1fbce)795962b)64ff0a7)37c1e4c)gix-diff (3fd9fab)af03832)gix-diff (5c1f010)93e86f1)012a754)d69c617)gix-features (ee80b2b)649f588)beba720)37ba461)2e0ce50)c3a7dcf)29ad2df)f992fb7)ec69c88)
</details>This release was made to supersede the borked gix v0.82 release, as it released gix-actor as patch-update which broke v0.81 and probably earlier versi
This release was made to supersede the borked gix v0.82 release, as it released gix-actor as patch-update which broke v0.81 and probably earlier versions as well.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
b3ff033)209b6de)
</details>A maintenance release without user-facing changes.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
4fe330e)f2b522d)a65a17f)d19af16)0f25841)
</details>assure duplicate packs and indices aren't overwriting existing ones. That way, on Windows there is no chance for access-denied errors. Maybe there are
<csr-id-a2da5373fada6c1a90d139bba6db7b238ae6504c/>
<csr-id-36d96f124bae7282da69883dc4c8e145261fb0f9/> assure duplicate packs and indices aren't overwriting existing ones. That way, on Windows there is no chance for access-denied errors. Maybe there are other advantages as well even for Unix.
Since packs are usually written rarely, in comparison to loose objects, the extra file accesss seems acceptable.
<csr-id-36d1b69ae6d8765936ebb0dc590a14704dc41e64/> do not write empty packs and indices
Failure can occour on Windows as it's likely such a pack or index is already opened
during negotiation. Then, when an empty fetch via --depth 1 is repeated,
a temporary file would be renamed onto one an mmapped pack or index, causing
a failure.
Now packs or indices aren't written anymore if they are empty.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 2 times to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
6232824)a1b73a6)0470df3)99c00cc)15f1cf7)bf3d82a)ab02aa9)a2da537)8d89b86)1dfa90d)tar files (dcab79a)9fa1054)9923542)ecfde07)f36b9bd)2f9f0ac)acc1331)9c65d98)ea12fc2)68cbea8)
</details>Progress report January 26 by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2393
gix-error conversions: gix-actor by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2396tracing instead of tracing-core to re-use the standard macros. by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2401core.bare being known or not by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2403blocking-http-transport-reqwest doesn't enable reqwest's default features by @fenhl in https://github.com/GitoxideLabs/gitoxide/pull/2404bytes to a fixed version by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2411gix-error by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2400imara-diff 0.2 in gix-blame by @cruessler in https://github.com/GitoxideLabs/gitoxide/pull/2420status work despite broken or invalid symlinks. by @Byron in https://github.com/GitoxideLabs/gitoxide/pull/2440Full Changelog: https://github.com/GitoxideLabs/gitoxide/compare/v0.50.0...v0.51.0
<csr-id-b32a847e10b742834169fac56c284645c1ed28f3/> don't panic when unknown entry types are encountered. Related to https://github.com/helix-editor/helix/issues/10660 which runs into object types that are unknown.
I have looked into this and couldn't find evidence of a new pack-entry type in the Git codebase.
It also looks like that Git will never write packs that aren't V2
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
4f98e94)9511416)04ef31e)gix-pack (bad5b48)b32a847)
</details>4 commits contributed to the release over the course of 12 calendar days.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
Exn::into_box() to Exn::into_inner(). (939b8fc)exn to most pressing needs of gitoxide (abedade)
</details>topo more similar to Ancestors, but also rename Ancestors to Simple<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
095c673)5755271)b590a9d)gix-traverse (1cfeb11)topo more similar to Ancestors, but also rename Ancestors to Simple (2a9c178)gix-traverse (6154bf3)7f6bee5)
</details>assure memory maps are created with MAP_PRIVATE That way, the mmap process should work under more circumstances.
<csr-id-39879af6eaf2bf4fe159a5c6371c98d516c4febe/>
MAP_PRIVATE
That way, the mmap process should work under more circumstances.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
MAP_PRIVATE (88061a1)b050327)52c3bbd)3e5c974)5722e3a)39879af)f8ce3d0)9e9b9fe)
</details>A maintenance release without user-facing changes.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
45b4470)f2e111f)13d5602)7b44c7f)8430442)684fa5c)2a9ef4e)d4d478b)2482023)0be338f)b8cba96)
</details>allow credential fill with gix credential fill to run without a repo
gix credential fill to run without a repoRemove doc_auto_cfg feature to fix docs.rs documentation.
It is part of doc_cfg feature since https://github.com/rust-lang/rust/pull/138907
This fixes the docs.rs build
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix credential fill to run without a repo (924cd62)doc_auto_cfg feature to fix docs.rs documentation. (6f469a6)
</details><csr-id-6f469a6fea59c88e6c69a5f94b0bc8a5977cb75b/>
gix credential fill to run without a repo<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
unsafe (4017e69)eb6aa8f)6a2e0be)b6c04c8)39f35da)3ef3bc2)2a663a0)c85a981)6587eea)ItemSliceMut::get_mut() bound-check to debug-code only. (911c05f)ef21617)
</details>3 commits contributed to the release.
<csr-id-3bd09ef120945a9669321ea856db4079a5dab930/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
972241f)8c492d7)rust-version manifest field back to 1.65. (3bd09ef)
</details>add a gix commit sign prototype
gix commit sign prototypegix branch list<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix commit sign prototype (aacc326)gix branch list (58b5060)gix tag list when JSON is requested (fb2766b)
</details>gix commit sign prototypegix branch list<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
8699d0f)82f788e)gix commit sign prototype (aacc326)4da2927)6cc8464)gix-blame (a721e74)gix-blame (268dac4)525873f)04650a7)6651548)e537611)gix branch list (58b5060)c149116)gix tag list when JSON is requested (fb2766b)
</details><csr-id-aea89c3ad52f1a800abb620e9a4701bdf904ff7d/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
e1aae19)e78a92b)4454c9d)aea89c3)gix-hash (859a092)1691ba6)7549559)35439de)98b08f4)
</details>add first debug version of gix tag list
add first debug version of gix tag list
gix revision list --long-hashes for faster iteration.
The performance of the short-hash generation was improved as well.
support for commitgraph list from..to to exercise the new 'hide' capability.
Enable precious file parsing in gix CLI by default, allow overrides.
That's pretty neat as one can now set GIX_PARSE_PRECIOUS=0 in the environment
to disable precious file parsing, good to see what difference it makes.
It's also possible to do this wiht gix -c gitoxide.parsePrecious=0.
add support for multiple blame ranges like gix blame -L <range> -L <other-range> ...
Update the blame subcommand to handle multiple line ranges. This allows specifying multiple -L options similar to the usage of git.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix tag list (37d3bf2)gix revision list --long-hashes for faster iteration. (ab52a49)gix-blame (4afc51d)commitgraph list from..to to exercise the new 'hide' capability. (c5bc49f)gix CLI by default, allow overrides. (1df1ebb)gix blame -L <range> -L <other-range> ... (36a6ffe)gix-blame (8143d69)
</details>gix-object::Find trait<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
d3fd11e)55d386a)d3dcbe5)51c7abc)115c993)Node an unsafe constructor (c4807de)Clone impl of ItemSliceSend and make it Sync instead (43bca11)Node private to resolve.rs (7301ca1)Node into resolve.rs (d655ee1)ItemSliceSend::get_mut() from Node::into_child_iter() (d27857d)Item-typed variable from node to item (d6b8977)f528ae8)f4c3380)5a0d93e)bb3224c)2b80d84)a0cc80d)613f018)4f5ab36)6e5ef3b)17a7c6a)c684b26)ff99a18)gix-object (203d69c)c8bd660)82b01c2)gix-object::Find trait (d6e7201)c0e72fb)b65a80b)
</details>This is a bit empty as release notes got lost in a series of accidents, but it looks like it wouldn't have been too informative anyway.
This is a bit empty as release notes got lost in a series of accidents, but it looks like it wouldn't have been too informative anyway.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
68e5432)1347a54)345712d)b842691)gix-features (9e7c3e1)
</details>Add missing command docs + small style fixups Changes that affect the help/usage text shown at runtime:
<csr-id-f952c101cc8686e685074e8604e0f332c06d6767/> Add missing command docs + small style fixups Changes that affect the help/usage text shown at runtime:
Add descriptions for a few gix commands that didn't have it:
gix is-clean, gix is-changed, and gix env.
Add a description of the debug-only ein panic command.
Small spacing and capitalization improvements for consistency.
Add a line break in a /// block for brevity of the top line.
Changes to code style that do not affect what is shown at runtime:
Add . at the end of some /// first lines. (This doesn't
affect runtime behavior because clap normalizes this away.)
Put /// above all #[...] in a few places where it wasn't.
Adjust comment wrapping in a couple of places for clarity.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
<csr-id-f952c101cc8686e685074e8604e0f332c06d6767/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
7891fb1)8a60d5b)
</details>add first 'debug' version of gix diff file
gix diff fileClippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix diff file (dbb1532)gix-blame (8a31d88)--since (330711d)gix-blame (e7084d8)<csr-id-ed327f6163f54756e58c20f86a563a97efb256ca/>
generate and streaming-input feature toggles.
That way, it's possible to not compile a bunch of code in gix if
the writing of packs isn't required.<csr-id-96a07e08e6090222cf398b46aa8d46b56f81f14d/> Use Vec::resize() instead of set_len()
Otherwise it's possible for uninitialized memory to be used as if it was initialized,
which can lead to strange behaviour.
As the buffer is re-used, it's not actually zeroing that much memory either.
<csr-id-24dd870919ba444aa8099c63a78ea120d47ec28e/> use prodash::Count to indicate that nothing more than counting is performed, in place of prodash::Progress
<csr-id-4b74996b19176cb0f00860b3db5a27819d63e7d0/> Make usage of decompression context explicit. That way, the context can be reused which is more efficient than recreating it from scratch for every little delta to decompress.
This leads to a performance gain of 1.3%.
dyn trait where possible.
This reduces compile time due to avoiding duplication.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
8bd0456)375db06)6135a5e)generate and streaming-input feature toggles. (0357b6c)dynification (f658fcc)dyn trait where possible. (072ee32)prodash::Count to indicate that nothing more than counting is performed, in place of prodash::Progress (24dd870)prodash (ed327f6)04494c6)9ad9c5b)Vec::resize() instead of set_len() (96a07e0)4b74996)363ee77)
</details>add gix env to print paths relevant to the Git installation.
gix blame -L start,endgix env to print paths relevant to the Git installation.gix blame to the CLI
That way it's possible to see the blame result of any file in the
repository.<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix blame -L start,end (4a78395). for consistency (ddddf02)gix env to print paths relevant to the Git installation. (75d689f)gix blame to the CLI (80e5804)
</details><csr-id-93feea269eebd114e866e6f29f4a73c0096df9e0/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
6c62e74)30b2761)f23ea88)8f3f358)73c685a)93feea2)4604f83)
</details>A maintenance release without user-facing changes.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
preserve SCP parsing alongside virtual drive prefixes
preserve SCP parsing alongside virtual drive prefixes
The Windows test-fast and fixture suites classified [::1]:repo and
[:[@]:... as local paths because their leading [: matched the broad
virtual-drive check. They also stopped rejecting an empty helper name in
::bar.
Require the drive prefix to end at the URL parser's selected colon. This
preserves the existing bracket-aware SCP interpretation and empty-helper
rejection while retaining digit and Unicode virtual-drive support.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
0062971)107a64e)16295b5)c4cc5f2)5cb3589)2fc66b5)9064ea3)
</details>add first 'debug' version of gix log It's primarily meant to better understand gix blame.
gix log
It's primarily meant to better understand gix blame.--tree-favor to gix merge tree|commit.
With it one can decide which side to favor in case of
irreconcilable tree-conflicts.gix merge file now uses THEIRS instead of OURS where needed<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix merge file now uses THEIRS instead of OURS where needed (0727b56)gix log (c7e04e9)gix-diff (960773e)--tree-favor to gix merge tree|commit. (471e046)A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
4aca8c2)68ae3ff)c70e54f)e4dded0)8a46a7e)33f95ba)3ef32af)b19a56d)3821089)
</details>A maintenance release without user-facing changes.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
69c6a36)c143cf4)gix-path across all crates (73c1292)
</details>add gix merge commit --debug This yields additional debugging information with details about detected conflicts.
gix merge commit --debug
This yields additional debugging information with details about detected conflicts.gix merge commitsgix merge tree to merge trees similarly to git merge-tree.<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix merge commit --debug (0a5993f)gix merge commits (4a5afc7)gix (8d590f3)gix merge tree to merge trees similarly to git merge-tree. (84707c2)add Url::path_query_fragment()
Expose the decoded path, query pairs, and fragment of Git remote URLs
separately. Consumers can identify repository paths and inspect HTTP(S)
parameters without implementing scheme-specific splitting and decoding.
Recognize delimiters before decoding and reuse validated original spelling
to borrow decoded bytes. Preserve encoded separators, nested escapes,
ordered duplicate query keys, and literal percent text after construction
or mutation. Query pairs use form decoding (+ becomes space); SSH/SCP
paths, existing accessors, and serialization keep their behavior.
Reference Git's connect.c::parse_connect_url() and url.c at
1630431f326e. Use fetch-pack --diag-url to check SSH delimiters;
standalone url-parse does not model that transport.
migrate errors to gix-error
raise MSRV to Rust 1.88
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
Url::path_query_fragment() (cd90fe1)A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
27e8c18)00f96fb)6fc7497)
</details>support for listing worktrees with gix worktree list
support for listing worktrees with gix worktree list
add first 'debug' version of gix diff tree.
It's primarily meant to better understand gix blame.
add new gix cat command.
It only prints things without fuzz.
Inspired by https://youtu.be/JYH5ILv5g1g?si=bHLBPFJiZyRUTl6u&t=211.
add gix merge-file with similar features as git merge-file
gix merge-base for the CLI
For now it only supports the standard merge-base, but more derivatives
can be added easily on demand.
gix clean warning and help for worktree fixgit clean options
-r/--repositories and --skip-hidden-repositories to make it
clearer what their relationship is, avoid creating the false
impression that repositories are never deleted in the absence of
-r/--repositories, and note that --skip-hidden-repositories
is sometimes needed to preserve not only separate repositories but
the (probably rarer) case of hidden nested worktrees of the current
repository.time to jiff
This swaps out time for jiff. It doesn't completely remove time
from the dependency tree. The last remaining use of time is in
prodash, outside of the gitoxide project.<csr-id-786bfec7c2cf25a68705b6ef09ba6390600742b8/> Unify style in config support info
This builds on b31d6b7 (#1531) by adjusting the capitalization and
quoting style of string arguments in progress.rs that help
document the status of what configuration options are and are not
planned.
Since I believe these strings are not usually rendered as Markdown, and most places that code formatting in Markdown would be used were already written with single quotes, this applies that quoting style consistently.
This applies initial capitalization, which was sometimes but not
always done. It does not apply a consistent style for when a
string will end in a ., which seems already to differ
deliberately for reasons of clarity. That can be adjusted later.
A small amount of minor rewording for clarity is also included.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix commit-graph is (49bca2a)gix diff tree. (0c554e0)gix cat command. (9c8bc03)gix merge-file with similar features as git merge-file (3da2da9)gix merge-base for the CLI (7249291)time to jiff (28ac657)gix clean warning and help for worktree fix (6eb4ea6)gix worktree list (c7213bc)
</details><csr-id-bcad5c22049d56a25ef69d6c7a3344e78f9a1d4d/>
tracing spans for common operations.
This is just the beginning and more crates will integrate with it over time.<csr-id-b9eb40730b53f788d2e4bffe4ef6d9028440782e/> revert 3a2d5286084597d4c68549903709cda77dda4357 to fix 'incorrect data check' error. This error could occour in heavily threaded code for unknown reason. But maybe it's due to threads somehow not cleaning up their reused decompressor properly (maybe related to the zlib-ng version). It's strange and sad as this really costs performnace for no good reason.
<csr-id-8817c248dd7c6453ced654d4df304f98ff18ecda/> don't crash when object validation failed during verification. When objects can't be serialized, they will trigger an error that manifests as IO error. Previously we didn't think of the possibility that writing to an im-memory buffer could fail would indeed panic during verification.
This is now fixed.
<csr-id-cd747f9292fb034e7fd5ee6c5a94a4e53df42e86/> ScopedJoinHandle is not always available
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
ea9f942)18b0a37)5a9a7a3)b9eb407)5861afb)8817c24)aa16c8c)4f635fc)tracing spans for common operations. (3cffa26)fe59956)cd747f9)clippy::redundant-closure-for-method-calls lint (bcad5c2)
</details>resolve Windows directory symlink chains
resolve Windows directory symlink chains
Delaying symlinks until regular files are written is insufficient on
Windows: an early link can target another link that has not been created
yet. A single target probe then leaves a file link where a directory link
is required, making paths such as a-child -> z-dir/nested unusable.
Git for Windows retries unresolved links when directory links appear.
Add gix_fs::symlink::create_or_delay() and retry unresolved checkout
entries while other links make progress. When a pass makes no progress,
create the remaining dangling or cyclic links with the existing file
fallback.
Completed links need no removal or recreation. For forced overwrites,
try creation before unlinking so a deferred replacement leaves the
existing link intact if a later object error aborts checkout.
cap automatic macOS checkout at four workers
Using all logical cores for checkout causes APFS creation contention on
macOS. Benchmark fresh checkouts of the Linux tree to choose a default
that preserves throughput without starting unnecessary workers.
Resolve None and Some(0) to at most four available logical cores in
shared checkout plumbing, so clone, linked-worktree creation, and index
checkout inherit the cap. Positive explicit limits remain overrides;
other platforms retain their existing defaults, and builds without
gix-parallel/parallel continue to run on the calling thread. Update the
checkout.workers and clone documentation accordingly.
On an Apple M4 Max with 16 logical CPUs and native case-insensitive APFS,
the initial worker sweep measured median checkout times of 2.564 s at
four workers, 3.774 s at eight, and 6.874 s at automatic sixteen. An initial
probe using newer offline-resolved dependencies favored five workers,
but that advantage did not reproduce with the workspace's locked versions.
The final comparison using those locked versions covers every execution
order of four, five, and six workers, with six samples per setting:
| Workers | Median seconds | Range |
|---|---|---|
| 4 | 2.664813 | 2.533386–3.106756 |
| 5 | 2.665293 | 2.513454–2.938505 |
| 6 | 2.903967 | 2.843629–4.047406 |
Four and five split the paired wins three to three and have effectively
identical medians. Choose four to obtain the same measured throughput
with one fewer worker, matching the existing macOS removal cap.
A separate three-round comparison with the workspace's locked dependencies
checks the same collision-safe Linux tree against Apple Git 2.54.0. Rotate
execution order so each method runs first, second, and third once:
| Checkout | Median seconds | Range |
|---|---|---|
| Git, one worker | 7.293862 | 7.279254–7.394098 |
gix, one thread |
6.875375 | 6.822443–6.958276 |
gix, four workers |
2.376388 | 2.349792–2.493034 |
Git runs checkout-index --all --prefix=<fresh-destination>/ with
checkout.workers=1 against the prebuilt retained index. Its timer includes
process startup and index loading; gix times the checkout call with an
already loaded index. Index preparation, volume synchronization, complete
Git content/mode/symlink validation, and cleanup are outside both timers.
These are initial-checkout references, not timings of full clone or branch
switch workflows. No builds or tests overlap the timed samples.
Raw wall seconds in execution order:
gix/1 6.822443, gix/4 2.376388.gix/4 2.493034, Git 7.279254, gix/1 6.958276.gix/1 6.875375, gix/4 2.349792, Git 7.293862.All nine checkouts pass Git content verification. This is a separate run
from the four/five/six comparison above; compare methods within each run.
A separate four-pair run compares Apple Git 2.54.0 with
checkout.workers=3 and checkout.workers=4 on the same retained tree,
using checkout-index --all --prefix=<fresh-destination>/ as above.
Alternate execution order between rounds:
| Git workers | Median seconds | Range |
|---|---|---|
| 3 | 4.261867 | 4.058227–4.361033 |
| 4 | 3.745698 | 3.611599–3.768478 |
Four workers win all four pairs; three offers no advantage here. A
separate warmup with GIT_TRACE2_EVENT confirms four checkout--worker
processes; exclude that traced 3.367775 s sample from the timing table.
All eight untraced checkouts pass Git content, mode, and symlink checks.
Volume synchronization, validation, and removal remain outside timers;
no probe compilation overlaps these measurements.
Raw wall seconds in execution order:
This later run measures Git's worker scaling. The earlier one-worker Git
and gix figures are reference timings from their own run, rather than
additional matched samples in this comparison.
Borrow objects read-only from the bare Linux repository at
/Users/byron/dev/git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux,
commit 707df3375124b51048233625a7e1c801e8c8a7fd. Use independent metadata
and fresh destinations under /private/tmp, omitting thirteen later
case-fold collisions that native APFS cannot represent. The retained tree
contains 88,821 entries, including 79 symlinks, and 1,509,611,994 blob bytes.
Time only checkout; synchronize the volume, verify content, executable
modes, and symlinks against isolated Git, and remove destinations outside
the timer. No builds or tests overlap timed samples. Keep the benchmark
method and decisive measurements in this commit message.
These warm-cache initial-checkout results cover one Mac and one volume.
They do not establish an optimum for every filesystem, repository, filter
pipeline, or incremental checkout. Explicit worker limits remain available
for those workloads; no cold-cache or durable-write benchmark was made.
mailmap check similar to git check-mailmap.
That way it becomes a little easier to maybe try out other special cases
in the wild.gix clone with --ref support.
--ref is similar to --branch, but was renamed as it also supports
tags for example.<csr-id-202f3e48a96b38d32b11d28449358c7d1f3546ff/> Make it easier to compile gitoxide as dynlib
And also optimize compilation time: By making plumbing and porcelain
as modules the lib.rs, they can be compiled after the rmeta for the
dependencies are generated.
For the uni.rs which uses both plumbing and porcelain, this would avoid
compiling these two modules twice.
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
mailmap check (3e08fa3)mailmap check similar to git check-mailmap. (2a56387)gix clone with --ref support. (9bf01e4)src/plumbing/main.rs (4596a39)preserve allocation failure sources
rubberstamp
Keep the original TryReserveError inside the OutOfMemory I/O error so structured classification can distinguish allocator failure without treating malformed stream metadata as corruption.
raise MSRV to Rust 1.88
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
2560a2c)298f3d7)fdee9a2)gix-traverse (1f682fd)
</details>This release contains fixes relevant to the safety of checkouts:
This release contains fixes relevant to the safety of checkouts:
checkout respects options for core.protectHFS and core.protectNTFS.
This also adds gitoxide.core.protectWindows as a way to enforce
additional restrictions that are usually only available on Windows.
Note that core.protectNFS is always enabled by default, just like
it is in Git.
more robustness in the face of a trampling-herd of threads loading a single index. The motivating example is here: https://github.com/praetorian-inc/noseyparker/issues/179
Previously, it was possible for a trampling herd of threads to consolidate the disk state. Most of them would be 'needs-init' threads which could notice that the initialization already happened, and just use that.
But a thread might be late for the party and somehow manages to not get any newly loaded index, and thus tries to consolidate with what's on disk again. Then it would again determine no change, and return nothing, causing the caller to abort and not find objects it should find because it wouldn't see the index that it should have seen.
The reason the thread got into this mess is that the 'is-load-ongoing' flagging was racy itself, so it would not wait for ongoing loads and just conclude nothing happened. An extra delay (by yielding) now assures it either seees the loading state and waits for it, sees the newly loaded indices.
Note that this issue can be reproduced with:
'./target/release/gix -r repo-with-one-pack -t10 --trace odb stats --extra-header-lookup'
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
core.protectHFS and core.protectNTFS. (886d6b5)Store::at() where possible (17fea2a)replace fs with gix-fs
gix-fs already provides the filesystem helpers and recursive walkers.
Keeping their duplicate implementation in gix-features adds unnecessary
filesystem dependencies to the foundational feature crate.
Remove the fs module and the walkdir and fs-read-dir features. Migrate
workspace callers to gix_fs, move traversal requests to gix-fs/walkdir,
and update the manifests, lockfile, traversal documentation, and feature
check script together.
fs with gix-fs (cf1ba49)<csr-id-3a2d5286084597d4c68549903709cda77dda4357/> improve performance by reducing the trashing of zlib::Inflate.
Previously, for every zlib inflate operation, we would allocate memory
due to the creation of a decompressor.
This is now avoided by reusing thread-local copies of such a decompressor, which is reused and reset instead.
<csr-id-14e7ea0217af8a04ed2b50ff7b13c28335c29022/> improve performance by avoiding zeroying buffers.
Previously we would use resize(new_len, 0) to resize buffers, even though
these values would then be overwritten (or the buffer isn't available).
Now we use set_len(new_len) after calling reserve to do the same, but safe
a memset.
<csr-id-977e135bfa00bdbc1a8f8324f85347ec9078c84f/> static linked list delta cache with memory cap Previously, the 64 slot big LRU cache for pack deltas didn't use any memory limit which could lead to memory exhaustion in the face of untypical, large objects.
Now we add a generous default limit to do better in such situations. It's worth noting though that that even without any cache, the working set of buffers to do delta resolution takes considerable memory, despite trying to keep it minimal.
Note that for bigger objects, the cache is now not used at all, which probably leads to terrible performance as not even the base object can be cached.
<csr-id-f89cbc675b0acc67322e289e7b3a17288b9eae27/> check for interrupt more often Previously when traversing a pack it could appear to hang as checks were only performed on chunk or base (of a delta-tree) level.
Now interrupt checks are performed more often to stop all work much quicker.
<csr-id-923692bcda698b45d3d1ad518b29f6d30b23fbc0/> memory capped hashmap as pack delta cache won't trash memory as much. Previously it would take a buffer from the free-list, copy data into it, and when exceeding the capacity loose it entirely. Now the freelist is handled correctly.
<csr-id-d22dd8fcc22e8dbe30524a1bdddc09bc841db341/> index-backed tree traversal with a form of work-stealing. When delta-trees are unbalanced, in pathological cases it's possible that that one thread ends up with more than half of the work. In this case it's required that it manages to spawn its own threads to parallelize the work it has.
<csr-id-0fa04bcbdf3102c5435e64cfef894a1bfc8d6e7b/> make current thread-count accessible in slice-workers. Threads started for working on an entry in a slice can now see the amount of threads left for use (and manipulate that variable) which effectively allows them to implement their own parallelization on top of the current one.
This is useful if there is there is very imbalanced work within the slice itself.
While at it, we also make consumer functions mutable as they exsit per thread.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
de1f6c2)d22dd8f)14e7ea0)0fa04bc)969cc77)977e135)f89cbc6)923692b)9a9fa96)just fmt (ffc1276)8f15cec)ae845de)9525ac8)d9d9bc0)zlib::Inflate. (3a2d528)420553a)89a8cfe)2578e57)dbf8aa1)2087032)3ef5c90)3eff567)9375cd7)facaaf6)78d28a7)996ba6b)2f275d5)gix-features (c48bbe3)4f879bf)
</details>add gix status --ignored support
add gix status --ignored support
add gix status --index-worktree-renames
This enables rename-tracking between worktree and index, something
that Git also doesn't do or doesn't do by default.
It is, however, available in git2.
gix status with submodule and rewrite support.
Submodule changes are now picked up as long as the submodule is
in the index.
Further, it's possible to enable rename-tracking between
index and worktree separately.
add gix is-clean|is-changed
It's a good way to compare the time it takes to run a full status
compared to a quick is-dirty check.
gix submodules list --dirty-suffix for dirty-information
This is a submodule-centric and greatly simplified way of obtaining
describe information with dirty-suffix.
Note that status information is also possible, but it seems
hard to display nicely, which this command isn't great at
in the first place.
add gix commit describe --dirty-suffix
That way a suffix will be added depending on the dirty-state of the repository.
gix status --format to communicate the current format is very simple. (23bea36)gix status --ignored support (84c74ff)gix status --index-worktree-renames (66e87cd)gix status with submodule and rewrite support. (61c002b)gix is-clean|is-changed (98b3680)gix submodules list --dirty-suffix for dirty-information (afd20ca)gix commit describe --dirty-suffix (58231b4)migrate errors to gix-error
raise MSRV to Rust 1.88
The newly published dua-core 3.3 release used by linked-worktree removal
requires Rust 1.88, so raise every workspace crate and the advertised badge
together.
Keep the MSRV checks buildable by selecting the latest sysinfo and rusqlite
release lines that support Rust 1.88.
A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
7ebc9f7)0135158)43ebaf2)
</details>add gix clean --patterns-for-entries|-m to help with wildcards.
gix clean --patterns-for-entries|-m to help with wildcards.gix clean<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
-n argument as no-op to make them more compatible. (8959b21)gix clean --patterns-for-entries|-m to help with wildcards. (9863d75)gix clean (e8597f3)<csr-id-b83ee366a3c65c717beb587ad809268f1c54b8ad/> Rename serde1 cargo feature to serde and use the weak-deps cargo capability.
With it it's possible to not automatically declare all optional dependencies externally visible
features, and thus re-use feature names that oterwise are also a crate name.
Previously I thought that serde1 is for future-proofing and supporting multiple serde versions
at the same time. However, it's most definitely a burden I wouldn't want anyway, so using
serde seems to be the way to go into the future.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
Clippy helped 1 time to make code idiomatic.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
serde1 cargo feature to serde and use the weak-deps cargo capability. (b83ee36)d7173b2)e4df557)91134a1)30a1a71)6ebd61e)14e64e7)69faad0)gix-object (efeb29d)8dbd0a6)cdef398)19338d9)b574a39)
</details>A maintenance release without any user-facing changes.
A maintenance release without any user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
5dc1f9f)3016a28)typos tool. (a69f873)
</details>A maintenance release without any user-facing changes.
A maintenance release without any user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
c1f1bfb)c66e298)3836cc0)01277a6)
</details>Add rev parse --reference. It's similar to git rev-parse --symbolic-full-name.
rev parse --reference.
It's similar to git rev-parse --symbolic-full-name.mark gix::interrupt::init_handler() as unsafe
The passed interrupt() argument will be called from a signal
handler, so that needs to be documented and the call sites need to
state that they fulfill the contract.
Thanks to @Manishearth for pointing this out.
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
rev parse --reference. (193ffcd)gix::interrupt::init_handler() as unsafe (59b8104)
</details>A maintenance release without user-facing changes.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
29a0870)e06f5f5)
</details>gix rev parse --format to provide different versions of the same content. This only applies to blobs, but allows to obtain different versions of the s
gix rev parse --format to provide different versions of the same content.
This only applies to blobs, but allows to obtain different versions of the same blob
like:
--trace mode, greatly increase message-buffer size.
That way, it's much less likely that messages will get lost
due to being overwritten before they can be displayed every
100ms or so.GITOXIDE_* environment variables to GIX_#<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
GITOXIDE_* environment variables to GIX_# (2189cee)gix rev parse --format to provide different versions of the same content. (cf51a4d)gix-diff (1706e23)diff::resource_cache() low-level utility for rapid in-memory diffing of combinations of resources. (4aea9b0)fsck connectivity into just fsck much like git fsck (7ab5c76)--tree-style raw. (8dfbb4b)-c options are given on the command-line. (f9ae1bc)--trace mode, greatly increase message-buffer size. (b230078)gix free discover to inform about repository discovery. (886289f)<csr-id-ce182d6dae7e277d104893f0eec3285709946233/> don't over-estimate the number of objects to be received in the indexing phase. Previously, it would work with a number that is pessimistically duplicated as each object could in theory be a ref-delta, which then causes another base to be added to the stream, duplicating the amount of objects we need to account for in the acceleration data structure, which unfortunatelly cannot actually grow dynamically without violating invariants.
Now we simply don't try to estimata the number of ref-deltas (which don't exist here anymore) at this stage.
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
0f411e9)gix-pack (6db30ef)ce182d6)
</details>This release includes an overhaul of the url parsing implementation, allowing more urls to be parsed correctly and refining the error messages for inv
This release includes an overhaul of the url parsing implementation, allowing more urls to be parsed correctly and refining the error messages for
invalid ones. Improvements include the added support for ssh aliases (github:byron/gitoxide has previously been parsed as local path), adjustments
around the interpretation of colons in file names (previously we disallowed colons that were not followed up with a slash character) and some smaller
changes that bring the interpretation of file urls more in line with Git's implementation.
There are still many (edge) cases in Git's url parsing implementation which are not handled correctly by Gitoxide. If you notice any such deviation please open a new issue to help us making Gitoxide even more correct.
gix status auto-writes changed indices.
This prevents expensive operations to re-occour.gix status -s/--statistics to obtain additional information on what happened.
This is useful for understanding performance characteristics in detail.<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
gix status auto-writes changed indices. (46e5919)gix status -s/--statistics to obtain additional information on what happened. (7ba2fa1)
</details><csr-id-3ba25202240d13fdda998581297616afe06422ca/>
<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
ea9fd1d)e6cc618)3cfbf89)addd789)94c99c7)d686d94)dashmap in favor of own sharded concurrent hashmap. (3ba2520)gix-tempfile (bfcd1e3)gix-features (a0ed614)gix-features, use process::count_with_decimals() (3968133)56c005b)
</details>compatibility with bstr v1.3, use *.as_bytes() instead of .as_ref(). as_ref() relies on a known target type which isn't always present. However, once
bstr v1.3, use *.as_bytes() instead of .as_ref().
as_ref() relies on a known target type which isn't always present. However, once
there is only one implementation, that's no problem, but when that changes compilation
fails due to ambiguity.<csr-read-only-do-not-edit/>
<csr-read-only-do-not-edit/>
<details><summary>view details</summary>
9604783)bstr v1.3, use *.as_bytes() instead of .as_ref(). (135d317)
</details>One column per quarter.