Junio C Hamano [Fri, 27 Feb 2026 23:11:52 +0000 (15:11 -0800)]
Merge branch 'ps/ci-gitlab-msvc-updates'
CI update.
* ps/ci-gitlab-msvc-updates:
gitlab-ci: handle failed tests on MSVC+Meson job
gitlab-ci: use "run-test-slice-meson.sh"
ci: make test slicing consistent across Meson/Make
github: fix Meson tests not executing at all
meson: fix MERGE_TOOL_DIR with "--no-bin-wrappers"
ci: don't skip smallest test slice in GitLab
ci: handle failures of test-slice helper
Junio C Hamano [Fri, 27 Feb 2026 23:11:52 +0000 (15:11 -0800)]
Merge branch 'jc/whitespace-incomplete-line'
It does not make much sense to apply the "incomplete-line"
whitespace rule to symbolic links, whose contents almost always
lack the final newline. "git apply" and "git diff" are now taught
to exclude them for a change to symbolic links.
* jc/whitespace-incomplete-line:
whitespace: symbolic links usually lack LF at the end
Junio C Hamano [Fri, 27 Feb 2026 23:11:51 +0000 (15:11 -0800)]
Merge branch 'jc/checkout-switch-restore'
"git switch <name>", in an attempt to create a local branch <name>
after a remote tracking branch of the same name gave an advise
message to disambiguate using "git checkout", which has been
updated to use "git switch".
* jc/checkout-switch-restore:
checkout: tell "parse_remote_branch" which command is calling it
checkout: pass program-readable token to unified "main"
Junio C Hamano [Fri, 27 Feb 2026 23:11:50 +0000 (15:11 -0800)]
Merge branch 'ps/history-ergonomics-updates'
UI improvements for "git history reword".
* ps/history-ergonomics-updates:
Documentation/git-history: document default for "--update-refs="
builtin/history: rename "--ref-action=" to "--update-refs="
builtin/history: replace "--ref-action=print" with "--dry-run"
builtin/history: check for merges before asking for user input
builtin/history: perform revwalk checks before asking for user input
Junio C Hamano [Fri, 27 Feb 2026 23:11:50 +0000 (15:11 -0800)]
Merge branch 'ps/for-each-ref-in-fixes'
A handful of places used refs_for_each_ref_in() API incorrectly,
which has been corrected.
* ps/for-each-ref-in-fixes:
bisect: simplify string_list memory handling
bisect: fix misuse of `refs_for_each_ref_in()`
pack-bitmap: fix bug with exact ref match in "pack.preferBitmapTips"
pack-bitmap: deduplicate logic to iterate over preferred bitmap tips
Junio C Hamano [Wed, 25 Feb 2026 19:54:18 +0000 (11:54 -0800)]
Merge branch 'mc/tr2-process-ancestry-cleanup'
Add process ancestry data to trace2 on macOS to match what we
already do on Linux and Windows. Also adjust the way Windows
implementation reports this information to match the other two.
* mc/tr2-process-ancestry-cleanup:
t0213: add trace2 cmd_ancestry tests
test-tool: extend trace2 helper with 400ancestry
trace2: emit cmd_ancestry data for Windows
trace2: refactor Windows process ancestry trace2 event
build: include procinfo.c impl for macOS
trace2: add macOS process ancestry tracing
Junio C Hamano [Wed, 25 Feb 2026 19:54:16 +0000 (11:54 -0800)]
Merge branch 'cc/lop-filter-auto'
"auto filter" logic for large-object promisor remote.
* cc/lop-filter-auto:
fetch-pack: wire up and enable auto filter logic
promisor-remote: change promisor_remote_reply()'s signature
promisor-remote: keep advertised filters in memory
list-objects-filter-options: support 'auto' mode for --filter
doc: fetch: document `--filter=<filter-spec>` option
fetch: make filter_options local to cmd_fetch()
clone: make filter_options local to cmd_clone()
promisor-remote: allow a client to store fields
promisor-remote: refactor initialising field lists
When a strip option to the %(refname) placeholder is asked to leave N
path components, we first count up the path components to know how many
to remove. That happens with a loop like this:
/* Find total no of '/' separated path-components */
for (i = 0; p[i]; p[i] == '/' ? i++ : *p++)
;
which is a little hard to understand for two reasons.
First, the dereference in "*p++" is seemingly useless, since nobody
looks at the result. And static analyzers like Coverity will complain
about that. But removing the "*" will cause gcc to complain with
-Wint-conversion, since the two sides of the ternary do not match (one
is a pointer and the other an int).
Second, it is not clear what the meaning of "p" is at each iteration of
the loop, as its position with respect to our walk over the string
depends on how many slashes we've seen. The answer is that by itself, it
doesn't really mean anything: "p + i" represents the current state of
our walk, with "i" counting up slashes, and "p" by itself essentially
meaningless.
None of this behaves incorrectly, but ultimately the loop is just
counting the slashes in the refname. We can do that much more simply
with a for-loop iterating over the string and a separate slash counter.
We can also drop the comment, which is somewhat misleading. We are
counting slashes, not components (and a comment later in the function
makes it clear that we must add one to compensate). In the new code it
is obvious that we are counting slashes here.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Thu, 19 Feb 2026 07:57:52 +0000 (08:57 +0100)]
bisect: simplify string_list memory handling
We declare the refs_for_removal string_list as NODUP, forcing us to
manually allocate strings we insert. And then when it comes time to
clean up, we set strdup_strings so that string_list_clear() will free
them for us.
This is a confusing pattern, and can be done much more simply by just
declaring the list with the DUP initializer in the first place.
It was written this way originally because one of the callsites
generated the item using xstrfmt(). But that spot switched to a plain
xstrdup() in the preceding commit. That means we can now just let the
string_list code handle allocation itself.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
All callers of `refs_for_each_ref_in()` pass in a string that is
terminated with a trailing slash to indicate that they only want to see
refs in that specific ref hierarchy. This is in fact a requirement if
one wants to use this function, as the function trims the prefix from
each yielded ref. So if there was a reference that was called
"refs/bisect" as in our example, the result after trimming would be the
empty string, and that's something we disallow.
Fix this by adding the trailing slash.
Furthermore, taking a closer look, we strip the prefix only to re-add it
in `mark_for_removal()`. This is somewhat roundabout, as we can instead
call `refs_for_each_fullref_in()` to not do any stripping at all. Do so
to simplify the code a bit.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
pack-bitmap: fix bug with exact ref match in "pack.preferBitmapTips"
The "pack.preferBitmapTips" configuration allows the user to specify
which references should be preferred when generating bitmaps. This
option is typically expected to be set to a reference prefix, like for
example "refs/heads/".
It's not unreasonable though for a user to configure one specific
reference as preferred. But if they do, they'll hit a `BUG()`:
$ git -c pack.preferBitmapTips=refs/heads/main repack -adb
BUG: ../refs/iterator.c:366: attempt to trim too many characters
error: pack-objects died of signal 6
The root cause for this bug is how we enumerate these references. We
call `refs_for_each_ref_in()`, which will:
- Yield all references that have a user-specified prefix.
- Trim each of these references so that the prefix is removed.
Typically, this function is called with a trailing slash, like
"refs/heads/", and in that case things work alright. But if the function
is called with the name of an existing reference then we'll try to trim
the full reference name, which would leave us with an empty name. And as
this would not really leave us with anything sensible, we call `BUG()`
instead of yielding this reference.
One could argue that this is a bug in `refs_for_each_ref_in()`. But the
question then becomes what the correct behaviour would be:
- Do we want to skip exact matches? In our case we certainly don't
want that, as the user has asked us to generate a bitmap for it.
- Do we want to yield the reference with the empty refname? That would
lead to a somewhat weird result.
Neither of these feel like viable options, so calling `BUG()` feels like
a sensible way out. The root cause ultimately is that we even try to
trim the whole refname in the first place. There are two possible ways
to fix this issue:
- We can fix the bug by using `refs_for_each_fullref_in()` instead,
which does not strip the prefix at all. Consequently, we would now
start to accept all references that start with the configured
prefix, including exact matches. So if we had "refs/heads/main", we
would both match "refs/heads/main" and "refs/heads/main-branch".
- Or we can fix the bug by appending a slash to the prefix if it
doesn't already have one. This would mean that we only match
ref hierarchies that start with this prefix.
While the first fix leaves the user with strictly _more_ configuration
options, we have already fixed a similar case in 10e8a9352b (refs.c:
stop matching non-directory prefixes in exclude patterns, 2025-03-06) by
using the second option. So for the sake of consistency, let's apply the
same fix here.
Clarify the documentation accordingly.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
pack-bitmap: deduplicate logic to iterate over preferred bitmap tips
We have two locations that iterate over the preferred bitmap tips as
configured by the user via "pack.preferBitmapTips". Both of these
callsites are subtly wrong: when the preferred bitmap tips contain an
exact refname match, then we will hit a `BUG()`.
Prepare for the fix by unifying the two callsites into a new
`for_each_preferred_bitmap_tip()` function.
This removes the last callsite of `bitmap_preferred_tips()` outside of
"pack-bitmap.c". As such, convert the function to be local to that file
only. Note that the function is still used by a second caller, so we
cannot just inline it.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The MSVC+Meson job does not currently have any logic to print failing
tests, nor does it upload the failed test artifacts. Backfill this logic
to make help debugging efforts in case any of its jobs has failed.
GitHub already knows to do this, so we don't need an equivalent change
over there.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
While our GitHub workflow already uses "ci/run-test-slice-meson.sh",
GitLab CI open-codes the parameters. Adapt the latter to also use the
same script so that we always use the same Meson options across both CI
systems.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
ci: make test slicing consistent across Meson/Make
In the preceding commit we have adjusted test slicing to be one-based
when using the "ci/run-test-slice.sh" script. But we also have an
equivalent script for Meson that is still zero-based, which is of course
inconsistent.
Adapt the script to be one-based, as well, and adapt the GitHub workflow
accordingly. Note that GitLab doesn't yet use the script, so it does not
need to be adapted. This will change in the next commit though.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
While the win+Meson test jobs run in GitHub workflows, the shell script
that is supposed to run the jobs is seemingly not running at all. All
that the CI job prints is the following:
The step is currently defined to use PowerShell, and of course it
doesn't know how to execute POSIX shell scripts. What's surprising
though is that this step doesn't even lead to a CI failure.
Fix the issue by using Bash instead of PowerShell, as we do in other
steps that execute shell scripts.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
meson: fix MERGE_TOOL_DIR with "--no-bin-wrappers"
On Windows, we execute tests with "--no-bin-wrappers". This has been
introduced via a87e427e35 (ci: speed up Windows phase, 2019-01-29) to
save some time: spawning processes is expensive on Windows, and shell
scripts tend to spawn a bunch of them. So overall, the bin-wrappers led
to a performance overhead of ~10-30%.
This causes test failures when using Meson on Windows:
failure: t7610.28 mergetool --tool-help shows recognized tools
++ git mergetool --tool-help
/d/a/git/git/build/git-mergetool--lib: line 45: cd: D:/a/git/git/build/mergetools: No such file or directory
The root cause here is that our bin-wrappers are usually responsible for
setting up the `MERGE_TOOL_DIR` environment variable so that we can
locate these scripts. But as we don't use the bin-wrappers, we'll
instead use the default location for merge tools, which is derived from
`GIT_EXEC_PATH`. And as `GIT_EXEC_PATH` points to our build directory,
which won't ever contain any of the merge tools, we will fail to locate
any of the merge tools.
This issue has went unnoticed for a long time given that we only skip
bin-wrappers on Windows, and because the CI jobs on Windows didn't
execute due to a bug.
Fix the issue by always setting the `MERGE_TOOL_DIR` environment
variable to the correct directory.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "ci/run-test-slice.sh" script can be used to slice up all of our
tests into N pieces and then run each of them on a separate CI job.
This is used by both GitLab and GitHub CI to speed up Windows tests,
which would otherwise be painfully slow.
The infra itself is fueled by `test-tool path-utils slice-tests`. This
tool receives as input an "offset" and a "stride" that can be combined
to slice up tests. This framing can be misleading though: you are
expected to pass a zero-based index as "offset", and the complete number
of slices to the "stride". The latter makes sense, but it is somewhat
surprising that the offset needs to be zero-based. And this is in fact
biting us: while GitHub passes zero-based indices, GitLab passes
`$CI_NODE_INDEX`, which is a one-based indice.
Ideally, we should have verification that the parameters make sense.
And naturally, one would for example expect that it's an error to call
the binary with an offset larger than the stride. But with the current
framing as "offset" it's not even wrong to do so, as it is of course
well-defined to start at a larger offset than the stride.
This means that we get this wrong on GitLab's CI, as we pass a one based
index there, and this causes us to skip one of the tests. Interestingly,
it's not the lexicographically first test that we skip. Instead, as we
sort tests by size before slicing them, we skip the _smallest_ test.
Reframe the problem to instead talk about "slice number" and "total
number of slices". For all of our use cases this is semantically
equivalent, but it allows us to perform some verifications:
- The total number of slices must be greater than 1.
- The selected slice must be between 1 <= nr <= slices_total.
As the indices are now one-based it means that GitLab's CI is fixed.
The GitHub workflow is updated accordingly.
Helped-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "run-test-slice.sh" script executes the test helper to slice up
tests passed to it. As the execution is part of a pipe though, we end up
ignoring any potential error code returned by the helper.
Make the code more robust by storing the tests in a variable first so
that we can split up the pipeline.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Tue, 17 Feb 2026 21:30:41 +0000 (13:30 -0800)]
Merge branch 'yt/merge-file-outside-a-repository'
"git merge-file" can be run outside a repository, but it ignored
all configuration, even the per-user ones. The command now uses
available configuration files to find its customization.
* yt/merge-file-outside-a-repository:
merge-file: honor merge.conflictStyle outside of a repository
Christian Couder [Mon, 16 Feb 2026 13:23:15 +0000 (14:23 +0100)]
fetch-pack: wire up and enable auto filter logic
Previous commits have set up an infrastructure for `--filter=auto` to
automatically prepare a partial clone filter based on what the server
advertised and the client accepted.
Using that infrastructure, let's now enable the `--filter=auto` option
in `git clone` and `git fetch` by setting `allow_auto_filter` to 1.
Note that these small changes mean that when `git clone --filter=auto`
or `git fetch --filter=auto` are used, "auto" is automatically saved
as the partial clone filter for the server on the client. Therefore
subsequent calls to `git fetch` on the client will automatically use
this "auto" mode even without `--filter=auto`.
Let's also set `allow_auto_filter` to 1 in `transport.c`, as the
transport layer must be able to accept the "auto" filter spec even if
the invoking command hasn't fully parsed it yet.
When an "auto" filter is requested, let's have the "fetch-pack.c" code
in `do_fetch_pack_v2()` compute a filter and send it to the server.
In `do_fetch_pack_v2()` the logic also needs to check for the
"promisor-remote" capability and call `promisor_remote_reply()` to
parse advertised remotes and populate the list of those accepted (and
their filters).
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `promisor_remote_reply()` function performs two tasks:
1. It uses filter_promisor_remote() to parse the server's
"promisor-remote" advertisement and to mark accepted remotes in the
repository configuration.
2. It assembles a reply string containing the accepted remote names to
send back to the server.
In a following commit, the fetch-pack logic will need to trigger the
side effect (1) to ensure the repository state is correct, but it will
not need to send a reply (2).
To avoid assembling a reply string when it is not needed, let's change
the signature of promisor_remote_reply(). It will now return `void` and
accept a second `char **accepted_out` argument. Only if that argument
is not NULL will a reply string be assembled and returned back to the
caller via that argument.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Christian Couder [Mon, 16 Feb 2026 13:23:13 +0000 (14:23 +0100)]
promisor-remote: keep advertised filters in memory
Currently, advertised filters are only kept in memory temporarily
during parsing, or persisted to disk if `promisor.storeFields`
contains 'partialCloneFilter'.
In a following commit though, we will add a `--filter=auto` option.
This option will enable the client to use the filters that the server
is suggesting for the promisor remotes the client accepts.
To use them even if `promisor.storeFields` is not configured, these
filters should be stored somewhere for the current session.
Let's add an `advertised_filter` field to `struct promisor_remote`
for that purpose.
To ensure that the filters are available in all cases,
filter_promisor_remote() captures them into a temporary list and
applies them to the `promisor_remote` structs after the potential
configuration reload.
Then the accepted remotes are marked as `accepted` in the repository
state. This ensures that subsequent calls to look up accepted remotes
(like in the filter construction below) actually find them.
In a following commit, we will add a `--filter=auto` option that will
enable a client to use the filters suggested by the server for the
promisor remotes the client accepted.
To enable the client to construct a filter spec based on these filters,
let's also add a `promisor_remote_construct_filter(repo)` function.
This function:
- iterates over all accepted promisor remotes in the repository,
- collects the filters advertised for them (using `advertised_filter`
added in this commit, and
- generates a single filter spec for them.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Christian Couder [Mon, 16 Feb 2026 13:23:12 +0000 (14:23 +0100)]
list-objects-filter-options: support 'auto' mode for --filter
In a following commit, we are going to allow passing "auto" as a
<filterspec> to the `--filter=<filterspec>` option, but only for some
commands. Other commands that support the `--filter=<filterspec>`
option should still die() when 'auto' is passed.
Let's set up the "list-objects-filter-options.{c,h}" infrastructure to
support that:
- Add a new `unsigned int allow_auto_filter : 1;` flag to
`struct list_objects_filter_options` which specifies if "auto" is
accepted or not by the current command.
- Change gently_parse_list_objects_filter() to parse "auto" if it's
accepted.
- Make sure we die() if "auto" is combined with another filter.
- Update list_objects_filter_release() to preserve the
allow_auto_filter flag, as this function is often called (via
opt_parse_list_objects_filter) to reset the struct before parsing a
new value.
Let's also update `list-objects-filter.c` to recognize the new
`LOFC_AUTO` choice. Since "auto" must be resolved to a concrete filter
before filtering actually begins, initializing a filter with
`LOFC_AUTO` is invalid and will trigger a BUG().
Note that ideally combining "auto" with "auto" could be allowed, but in
practice, it's probably not worth the added code complexity. And if we
really want it, nothing prevents us to allow it in future work.
If we ever want to give a meaning to combining "auto" with a different
filter too, nothing prevents us to do that in future work either.
Also note that the new `allow_auto_filter` flag depends on the command,
not user choices, so it should be reset to the command default when
`struct list_objects_filter_options` instances are reset.
While at it, let's add a new "u-list-objects-filter-options.c" file for
`struct list_objects_filter_options` related unit tests. For now it
only tests gently_parse_list_objects_filter() though.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Christian Couder [Mon, 16 Feb 2026 13:23:10 +0000 (14:23 +0100)]
fetch: make filter_options local to cmd_fetch()
The `struct list_objects_filter_options filter_options` variable used
in "builtin/fetch.c" to store the parsed filters specified by
`--filter=<filterspec>` is currently a static variable global to the
file.
As we are going to use it more in a following commit, it could become a
bit less easy to understand how it's managed.
To avoid that, let's make it clear that it's owned by cmd_fetch() by
moving its definition into that function and making it non-static.
This requires passing a pointer to it through the prepare_transport(),
do_fetch(), backfill_tags(), fetch_one_setup_partial(), and fetch_one()
functions, but it's quite straightforward.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Christian Couder [Mon, 16 Feb 2026 13:23:09 +0000 (14:23 +0100)]
clone: make filter_options local to cmd_clone()
The `struct list_objects_filter_options filter_options` variable used
in "builtin/clone.c" to store the parsed filters specified by
`--filter=<filterspec>` is currently a static variable global to the
file.
As we are going to use it more in a following commit, it could become
a bit less easy to understand how it's managed.
To avoid that, let's make it clear that it's owned by cmd_clone() by
moving its definition into that function and making it non-static.
The only additional change to make this work is to pass it as an
argument to checkout(). So it's a small quite cheap cleanup anyway.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Christian Couder [Mon, 16 Feb 2026 13:23:08 +0000 (14:23 +0100)]
promisor-remote: allow a client to store fields
A previous commit allowed a server to pass additional fields through
the "promisor-remote" protocol capability after the "name" and "url"
fields, specifically the "partialCloneFilter" and "token" fields.
Another previous commit, c213820c51 (promisor-remote: allow a client
to check fields, 2025-09-08), has made it possible for a client to
decide if it accepts a promisor remote advertised by a server based
on these additional fields.
Often though, it would be interesting for the client to just store in
its configuration files these additional fields passed by the server,
so that it can use them when needed.
For example if a token is necessary to access a promisor remote, that
token could be updated frequently only on the server side and then
passed to all the clients through the "promisor-remote" capability,
avoiding the need to update it on all the clients manually.
Storing the token on the client side makes sure that the token is
available when the client needs to access the promisor remotes for a
lazy fetch.
To allow this, let's introduce a new "promisor.storeFields"
configuration variable.
Note that for a partial clone filter, it's less interesting to have
it stored on the client. This is because a filter should be used
right away and we already pass a `--filter=<filter-spec>` option to
`git clone` when starting a partial clone. Storing the filter could
perhaps still be interesting for information purposes.
Like "promisor.checkFields" and "promisor.sendFields", the new
configuration variable should contain a comma or space separated list
of field names. Only the "partialCloneFilter" and "token" field names
are supported for now.
When a server advertises a promisor remote, for example "foo", along
with for example "token=XXXXX" to a client, and on the client side
"promisor.storeFields" contains "token", then the client will store
XXXXX for the "remote.foo.token" variable in its configuration file
and reload its configuration so it can immediately use this new
configuration variable.
A message is emitted on stderr to warn users when the config is
changed.
Note that even if "promisor.acceptFromServer" is set to "all", a
promisor remote has to be already configured on the client side for
some of its config to be changed. In any case no new remote is
configured and no new URL is stored.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Sun, 15 Feb 2026 09:07:44 +0000 (04:07 -0500)]
ref-filter: avoid strrchr() in rstrip_ref_components()
To strip path components from our refname string, we repeatedly call
strrchr() to find the trailing slash, shortening the string each time by
assigning NUL over it. This has two downsides:
1. Calling strrchr() in a loop is quadratic, since each call has to
call strlen() under the hood to find the end of the string (even
though we know exactly where it is from the last loop iteration).
2. We need a temporary buffer, since we're munging the string with NUL
as we shorten it (which we must do, because strrchr() has no other
way of knowing what we consider the end of the string).
Using memrchr() would let us fix both of these, but it isn't portable.
So instead, let's just open-code the string traversal from back to
front as we loop.
I doubt that the quadratic nature is a serious concern. You can see it
in practice with something like:
That takes ~5.5s to run on my machine before this patch, and ~11ms
after. But I don't think there's a reasonable way for somebody to infect
you with such a garbage ref, as the wire protocol is limited to 64k
pkt-lines. The difference is measurable for me for a 32k-component ref
(about 19ms vs 7ms), so perhaps you could create some chaos by pushing a
lot of them. But we also run into filesystem limits (if the loose
backend is in use), and in practice it seems like there are probably
simpler and more effective ways to waste CPU.
Likewise the extra allocation probably isn't really measurable. In fact,
since our goal is to return an allocated string, we end up having to
make the same allocation anyway (though it is sized to the result,
rather than the input). My main goal was simplicity in avoiding the need
to handle cleaning it up in the early return path.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're stripping path components from the end of a string, which we do by
assigning a NUL as we parse each component, shortening the string. This
requires an extra temporary buffer to avoid munging our input string.
But the way that we allocate the buffer is unusual. We have an extra
"to_free" variable. Usually this is used when the access variable is
conceptually const, like:
But that's not what's happening here. Our "start" variable always points
to the allocated buffer, and to_free is redundant. Worse, it is marked
as const itself, requiring a cast when we free it.
Let's drop to_free entirely, and mark "start" as non-const, making the
memory handling more clear. As a bonus, this also silences a warning
from glibc-2.43 that our call to strrchr() implicitly strips away the
const-ness of "start".
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're walking forward in the string, skipping path components from
left-to-right. So when we've stripped as much as we want, the pointer we
have is a complete NUL-terminated string and we can just return it
(after duplicating it, of course). So there is no need for a temporary
allocated string.
But we do make an extra temporary copy due to f0062d3b74 (ref-filter:
free item->value and item->value->s, 2018-10-18). This is probably from
cargo-culting the technique used in rstrip_ref_components(), which
_does_ need a separate string (since it is stripping from the end and
ties off the temporary string with a NUL).
Let's drop the extra allocation. This is slightly more efficient, but
more importantly makes the code much simpler.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Sun, 15 Feb 2026 09:00:52 +0000 (04:00 -0500)]
ref-filter: factor out refname component counting
The "lstrip" and "rstrip" options to the %(refname) placeholder both
accept a negative length, which asks us to keep that many path
components (rather than stripping that many).
The code to count components and convert the negative value to a
positive was copied from lstrip to rstrip in 1a34728e6b (ref-filter: add
an 'rstrip=<N>' option to atoms which deal with refnames, 2017-01-10).
Let's factor it out into a separate function. This reduces duplication
and also makes the lstrip/rstrip functions much easier to follow, since
the bulk of their code is now the actual stripping.
Note that the computed "remaining" value is currently stored as a
"long", so in theory that's what our function should return. But this is
purely historical. When the variable was added in 0571979bd6 (tag: do
not show ambiguous tag names as "tags/foo", 2016-01-25), we parsed the
value from strtol(), and thus used a long. But these days we take "len"
as an int, and also use an int to count up components. So let's just
consistently use int here. This value could only overflow in a
pathological case (e.g., 4GB worth of "a/a/...") and even then will not
result in out-of-bounds memory access (we keep stripping until we run
out of string to parse).
The minimal Myers diff here is a little hard to read; with --patience
the code movement is shown much more clearly.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-history: document default for "--update-refs="
While we document the values that can be passed to the "--update-refs="
option, we don't give the user any hint what the default behaviour is.
Document it.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin/history: rename "--ref-action=" to "--update-refs="
With the preceding commit we have changed "--ref-action=" to only
control which refs are supposed to be updated, not what happens with
them. As a consequence, the option is now somewhat misnamed, as we don't
control the action itself anymore.
Rename it to "--update-refs=" to better align it with its new use.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin/history: replace "--ref-action=print" with "--dry-run"
The git-history(1) command has the ability to perform a dry-run
that will not end up modifying any references. Instead, we'll only print
any ref updates that would happen as a consequence of performing the
operation.
This mode is somewhat hidden though behind the "--ref-action=print"
option. This command line option has its origin in git-replay(1), where
it's probably an okayish interface as this command is sitting more on
the plumbing side of tools. But git-history(1) is a user-facing tool,
and this way of achieving a dry-run is way too technical and thus not
very discoverable.
Besides usability issues, it also has another issue: the dry-run mode
will always operate as if the user wanted to rewrite all branches. But
in fact, the user also has the option to only update the HEAD reference,
and they might want to perform a dry-run of such an operation, too. We
could of course introduce "--ref-action=print-head", but that would
become even less ergonomic.
Replace "--ref-action=print" with a new "--dry-run" toggle. This new
toggle works with both "--ref-action={head,branches}" and is way more
discoverable.
Add a test to verify that both "--ref-action=" values behave as
expected.
This patch is best viewed with "--ignore-space-change".
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin/history: check for merges before asking for user input
The replay infrastructure is not yet capable of replaying merge commits.
Unfortunately, we only notice that we're about to replay merges after we
have already asked the user for input, so any commit message that the
user may have written will be discarded in that case.
Fix this by checking whether the revwalk contains merge commits before
we ask for user input.
Adapt one of the tests that is expected to fail because of this check
to use false(1) as editor. If the editor had been executed by Git, it
would fail with the error message "Aborting commit as launching the
editor failed."
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin/history: perform revwalk checks before asking for user input
When setting up the revision walk in git-history(1) we also perform some
verifications whether the request actually looks sane. Unfortunately,
these verifications come _after_ we have already asked the user for the
commit message of the commit that is to be rewritten. So in case any of
the verifications fails, the user will have lost their modifications.
Extract the function to set up the revision walk and call it before we
ask for user input to fix this.
Adapt one of the tests that is expected to fail because of this check
to use false(1) as editor. If the editor had been executed by Git, it
would fail with the error message "Aborting commit as launching the
editor failed."
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the user wants to find what are the available keys, they need to
either check the documentation or to ask for all the key-value pairs
by using --all.
Add a new flag --keys for listing only the available keys without
listing the values.
Signed-off-by: Lucas Seiki Oshiro <lucasseikioshiro@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
repo: rename the output format "keyvalue" to "lines"
Both subcommands in git-repo(1) accept the "keyvalue" format. This
format is newline-delimited, where the key is separated from the
value with an equals sign.
The name of this option is suboptimal though, as it is both too
limiting while at the same time not really indicating what it
actually does:
- There is no mention of the format being newline-delimited, which
is the key differentiator to the "nul" format.
- Both "nul" and "keyvalue" have a key and a value, so the latter
is not exactly giving any hint what makes it so special.
- "keyvalue" requires there to be, well, a key and a value, but we
want to add additional output that is only going to be newline
delimited.
Taken together, "keyvalue" is kind of a bad name for this output
format.
Luckily, the git-repo(1) command is still rather new and marked as
experimental, so things aren't cast into stone yet. Rename the
format to "lines" instead to better indicate that the major
difference is that we'll get newline-delimited output. This new name
will also be a better fit for a subsequent extension in git-repo(1).
Helped-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Lucas Seiki Oshiro <lucasseikioshiro@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Thu, 12 Feb 2026 21:22:56 +0000 (13:22 -0800)]
CodingGuidelines: document NEEDSWORK comments
We often say things like /* NEEDSWORK: further _do_ _this_ */ in
comments, but it is a short-hand to say "We might later want to do
this. We might not. We do not have to decide it right now at this
moment in the commit this comment was added. If somebody is
inclined to work in this area further, the first thing they need to
do is to figure out if it truly makes sense to do so, before blindly
doing it."
This seems to have never been documented. Do so now.
Junio C Hamano [Fri, 13 Feb 2026 21:39:26 +0000 (13:39 -0800)]
Merge branch 'jc/ci-test-contrib-too'
Test contrib/ things in CI to catch breakages before they enter the
"next" branch.
* jc/ci-test-contrib-too:
: Some of our downstream folks run more tests than we do and catch
: breakages in them, namely, where contrib/*/Makefile has "test" target.
: Let's make sure we fail upon accepting a new topic that break them in
: 'seen'.
ci: ubuntu: use GNU coreutils for dirname
test: optionally test contrib in CI
Junio C Hamano [Fri, 13 Feb 2026 21:39:25 +0000 (13:39 -0800)]
Merge branch 'jt/odb-transaction-per-source'
Transaction to create objects (or not) is currently tied to the
repository, but in the future a repository can have multiple object
sources, which may have different transaction mechanisms. Make the
odb transaction API per object source.
* jt/odb-transaction-per-source:
odb: transparently handle common transaction behavior
odb: prepare `struct odb_transaction` to become generic
object-file: rename transaction functions
odb: store ODB source in `struct odb_transaction`
Junio C Hamano [Fri, 13 Feb 2026 21:39:25 +0000 (13:39 -0800)]
Merge branch 'ps/commit-list-functions-renamed'
Rename three functions around the commit_list data structure.
* ps/commit-list-functions-renamed:
commit: rename `free_commit_list()` to conform to coding guidelines
commit: rename `reverse_commit_list()` to conform to coding guidelines
commit: rename `copy_commit_list()` to conform to coding guidelines
Junio C Hamano [Fri, 13 Feb 2026 21:39:25 +0000 (13:39 -0800)]
Merge branch 'tc/last-modified-not-a-tree'
Giving "git last-modified" a tree (not a commit-ish) died an
uncontrolled death, which has been corrected.
* tc/last-modified-not-a-tree:
last-modified: verify revision argument is a commit-ish
last-modified: remove double error message
last-modified: fix memory leak when more than one commit is given
last-modified: rewrite error message when more than one commit given
ISO C23 redefines strchr and friends that tradiotionally took
a const pointer and returned a non-const pointer derived from it to
preserve constness (i.e., if you ask for a substring in a const
string, you get a const pointer to the substring). Update code
paths that used non-const pointer to receive their results that did
not have to be non-const to adjust.
* cf/c23-const-preserving-strchr-updates-0:
gpg-interface: remove an unnecessary NULL initialization
global: constify some pointers that are not written to
Add a new test script t0213-trace2-ancestry.sh that verifies
cmd_ancestry events across all three trace2 output formats (normal,
perf, and event).
The tests use the "400ancestry" test helper to spawn child processes
with controlled trace2 environments. Git alias resolution (which
spawns a child git process) creates a predictable multi-level process
tree. Filter functions extract cmd_ancestry events from each format,
truncating the ancestor list at the outermost "test-tool" so that only
the controlled portion of the tree is verified, regardless of the test
runner environment.
A runtime prerequisite (TRACE2_ANCESTRY) is used to detect whether the
platform has a real procinfo implementation; platforms with only the
stub are skipped.
We must pay attention to an extra ancestor on Windows (MINGW) when
running without the bin-wrappers (such as we do in CI). In this
situation we see an extra "sh.exe" ancestor after "test-tool.exe".
Also update the comment in t0210-trace2-normal.sh to reflect that
ancestry testing now has its own dedicated test script.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a new test helper "400ancestry" to the trace2 test-tool that
spawns a child process with a controlled trace2 environment, capturing
only the child's trace2 output (including cmd_ancestry events) in
isolation.
The helper clears all inherited GIT_TRACE2* variables in the child
and enables only the requested target (normal, perf, or event),
directing output to a specified file. This gives the test suite a
reliable way to capture cmd_ancestry events: the child always sees
"test-tool" as its immediate parent in the process ancestry, providing
a predictable value to verify in tests.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 2f732bf15e (tr2: log parent process name, 2021-07-21) it is now
now possible to emit a specific process ancestry event in TRACE2. We
should emit the Windows process ancestry data with the correct event
type.
To not break existing consumers of the data_json "windows/ancestry"
event, we continue to emit the ancestry data as a JSON event.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
trace2: refactor Windows process ancestry trace2 event
In 353d3d77f4 (trace2: collect Windows-specific process information,
2019-02-22) we added process ancestry information for Windows to TRACE2
via a data_json event. It was only later in 2f732bf15e (tr2: log parent
process name, 2021-07-21) that the specific cmd_ancestry event was
added to TRACE2.
In a future commit we will emit the ancestry information with the newer
cmd_ancestry TRACE2 event. Right now, we rework this implementation of
trace2_collect_process_info to separate the calculation of ancestors
from building and emiting the JSON array via a data_json event.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In 353d3d77f4 (trace2: collect Windows-specific process information,
2019-02-22) Windows-specific process ancestry information was added as
a data_json event to TRACE2. Furthermore in 2f732bf15e (tr2: log
parent process name, 2021-07-21) similar functionality was added for
Linux-based systems, using procfs.
Teach Git to also log process ancestry on macOS using the sysctl with
KERN_PROC to get process information (PPID and process name).
Like the Linux implementation, we use the cmd_ancestry TRACE2 event
rather than using a data_json event and creating another custom data
point.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the body of a commit message contains a diff that is not indented
then "git am" will treat that diff as part of the patch rather than
as part of the commit message. This allows it to apply email messages
that were created by adding a commit message in front of a regular diff
without adding the "---" separator used by "git format-patch". This
often surprises users [1-4] so add a check to the sample "commit-msg"
hook to reject messages that would confuse "git am". Even if a project
does not use an email based workflow it is not uncommon for people
to generate patches from it and apply them with "git am". Therefore
it is still worth discouraging the creation of commit messages that
would not be applied correctly.
A further source of confusion when applying patches with "git am" is
the "---" separator that is added by "git format patch". If a commit
message body contains that line then it will be truncated by "git am".
As this is often used by patch authors to add some commentary that
they do not want to end up in the commit message when the patch is
applied, the hook does not complain about the presence of "---" lines
in the message.
Detecting if the message contains a diff is complicated by the
hook being passed the message before it is cleaned up so we need to
ignore any diffs below the scissors line. There are also two possible
config keys to check to find the comment character at the start of
the scissors line. The first paragraph of the commit message becomes
the email subject header which beings "Subject: " and so does not
need to be checked. The trailing ".*" when matching commented lines
ensures that if the comment string ends with a "$" it is not treated
as an anchor.
Phillip Wood [Fri, 13 Feb 2026 14:34:48 +0000 (14:34 +0000)]
templates: add .gitattributes entry for sample hooks
The sample hooks are shell scripts but the filenames end with ".sample"
so they need their own .gitattributes rule. Update our editorconfig
settings to match the attributes as well.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-format-patch(1) and git-am(1) deal with formatting commits as
patches and applying them, respectively. Naturally they use a few
delimiters to mark where the commit message ends. This can lead to
surprising behavior when these delimiters are used in the commit
message itself.
git-format-patch(1) will accept any commit message and not warn or error
about these delimiters being used.[1]
Especially problematic is the presence of unindented diffs in the commit
message; the patch machinery will naturally (since the commit message
has ended) try to apply that diff and everything after it.[2]
It is unclear whether any commands in this chain will learn to warn
about this. One concern could be that users have learned to rely on
the three-dash line rule to conveniently add extra-commit message
information in the commit message, knowing that git-am(1) will
ignore it.[4]
All of this is covered already, technically. However, we should spell
out the implications.
† 1: There is also git-commit(1) to consider. However, making that
command warn or error out over such delimiters would be disruptive
to all Git users who never use email in their workflow.
† 2: Recently patch(1) caused this issue for a project, but it was noted
that git-am(1) has the same behavior[3]
† 3: https://github.com/i3/i3/pull/6564#issuecomment-3858381425
† 4: https://lore.kernel.org/git/xmqqldh4b5y2.fsf@gitster.g/
https://lore.kernel.org/git/V3_format-patch_caveats.354@msgid.xyz/
Reported-by: Matthias Beyer <mail@beyermatthias.de> Reported-by: Christoph Anton Mitterer <calestyo@scientia.org> Reported-by: Matheus Tavares <matheus.tavb@gmail.com> Reported-by: Chris Packham <judge.packham@gmail.com> Helped-by: Jakob Haufe <sur5r@sur5r.net> Helped-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Kristoffer Haugsbakk <code@khaugsbakk.name> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Phillip Wood [Thu, 12 Feb 2026 15:53:50 +0000 (15:53 +0000)]
diff --anchored: avoid checking unmatched lines
For a line to be an anchor it has to appear in each of the files being
diffed exactly once. With that in mind lets delay checking whether
a line is an anchor until we know there is exactly one instance of
the line in each file. As each line is checked at most once, there
is no need to cache the result of is_anchor() and we can drop that
field from the hashmap entries. When diffing 5000 recent commits in
git.git this gives a modest speedup of ~2%. In the (rather extreme)
example below that consists largely of deletions the speedup is ~16%.
seq 0 10000000 >old
printf '%s\n' 300000 100000 200000 >new
git diff --no-index --anchored=300000 old new
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Wed, 11 Feb 2026 19:17:48 +0000 (11:17 -0800)]
CodingGuidelines: document // comments
We do not use // comments in our C code, which is implied by the
description of multi-line comment rule and its examples, but is not
explicitly spelled out. Spell it out.
Junio C Hamano [Wed, 11 Feb 2026 20:29:06 +0000 (12:29 -0800)]
Merge branch 'sp/show-index-warn-fallback'
When "git show-index" is run outside a repository, it silently
defaults to SHA-1; the tool now warns when this happens.
* sp/show-index-warn-fallback:
show-index: use gettext wrapping in user facing error messages
show-index: warn when falling back to SHA-1 outside a repository
builtin/pack-objects: don't fetch objects when merging packs
The "--stdin-packs" option can be used to merge objects from multiple
packfiles given via stdin into a new packfile. One big upside of this
option is that we don't have to perform a complete rev walk to enumerate
objects. Instead, we can simply enumerate all objects that are part of
the specified packfiles, which can be significantly faster in very large
repositories.
There is one downside though: when we don't perform a rev walk we also
don't have a good way to learn about the respective object's names. As a
consequence, we cannot use the name hashes as a heuristic to get better
delta selection.
We try to offset this downside though by performing a localized rev
walk: we queue all objects that we're about to repack as interesting,
and all objects from excluded packfiles as uninteresting. We then
perform a best-effort rev walk that allows us to fill in object names.
There is one gotcha here though: when "--exclude-promisor-objects" has
not been given we will perform backfill fetches for any promised objects
that are missing. This used to not be an issue though as this option was
mutually exclusive with "--stdin-packs". But that has changed recently,
and starting with dcc9c7ef47 (builtin/repack: handle promisor packs with
geometric repacking, 2026-01-05) we will now repack promisor packs
during geometric compaction. The consequence is that a geometric repack
may now perform a bunch of backfill fetches.
We of course cannot pass "--exclude-promisor-objects" to fix this
issue -- after all, the whole intent is to repack objects part of a
promisor pack. But arguably we don't have to: the rev walk is intended
as best effort, and we already configure it to ignore missing links to
other objects. So we can adapt the walk to unconditionally disable
fetching any missing objects.
Do so and add a test that verifies we don't backfill any objects.
Reported-by: Lukas Wanko <lwanko@gitlab.com> Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
projects / rit / log