refs: pass refname when invoking reflog entry callback
With `refs_for_each_reflog_ent()` callers can iterate through all the
reflog entries for a given reference. The callback that is being invoked
for each such entry does not receive the name of the reference that we
are currently iterating through. This isn't really a limiting factor, as
callers can simply pass the name via the callback data.
But this layout sometimes does make for a bit of an awkward calling
pattern. One example: when iterating through all reflogs, and for each
reflog we iterate through all refnames, we have to do some extra book
keeping to track which reference name we are currently yielding reflog
entries for.
Change the signature of the callback function so that the reference name
of the reflog gets passed through to it. Adapt callers accordingly and
start using the new parameter in trivial cases. The next commit will
refactor the reference migration logic to make use of this parameter so
that we can simplify its logic a bit.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Wed, 6 Aug 2025 14:41:57 +0000 (07:41 -0700)]
Merge branch 'ps/reflog-migrate-fixes' into ps/remote-rename-fix
* ps/reflog-migrate-fixes:
refs: fix invalid old object IDs when migrating reflogs
refs: stop unsetting REF_HAVE_OLD for log-only updates
refs/files: detect race when generating reflog entry for HEAD
refs: fix identity for migrated reflogs
ident: fix type of string length parameter
builtin/reflog: implement subcommand to write new entries
refs: export `ref_transaction_update_reflog()`
builtin/reflog: improve grouping of subcommands
Documentation/git-reflog: convert to use synopsis type
refs: fix invalid old object IDs when migrating reflogs
When migrating reflog entries between different storage formats we end
up with invalid old object IDs for the migrated entries: instead of
writing the old object ID of the to-be-migrated entry, we end up with
the all-zeroes object ID.
The root cause of this issue is that we don't know to use the old object
ID provided by the caller. Instead, we manually resolve the old object
ID by resolving the current value of its matching reference. But as that
reference does not yet exist in the target ref storage we always end up
resolving it to all-zeroes.
This issue got unnoticed as there is no user-facing command that would
even show the old object ID. While `git log -g` knows to show the new
object ID, we don't have any formatting directive to show the old object
ID.
Fix the bug by introducing a new flag `REF_LOG_USE_PROVIDED_OIDS`. If
set, backends are instructed to use the old and new object IDs provided
by the caller, without doing any manual resolving. Set this flag in
`ref_transaction_update_reflog()`.
Amend our tests in t1460-refs-migrate to use our test tool to read
reflog entries. This test tool prints out both old and new object ID of
each reflog entry, which fixes the test gap. Furthermore it also prints
the full identity used to write the reflog, which provides test coverage
for the previous commit in this patch series that fixed the identity for
migrated reflogs.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
refs: stop unsetting REF_HAVE_OLD for log-only updates
The `REF_HAVE_OLD` flag indicates whether a given ref update has its old
object ID set. If so, the value of that field is used to verify whether
the current state of the reference matches this expected state. It is
thus an important part of mitigating races with a concurrent process
that updates the same set of references.
When writing reflogs though we explicitly unset that flag. This is a
sensible thing to do: the old state of reflog entry updates may not
necessarily match the current on-disk state of its accompanying ref, but
it's only intended to signal what old object ID we want to write into
the new reflog entry. For example when migrating refs we end up writing
many reflog entries for a single reference, and most likely those reflog
entries will have many different old object IDs.
But unsetting this flag also removes a useful signal, namely that the
caller _did_ provide an old object ID for a given reflog entry. This
signal will become useful in a subsequent commit, where we add a new
flag that tells the transaction to use the provided old and new object
IDs to write a reflog entry. The `REF_HAVE_OLD` flag is then used as a
signal to verify that the caller really did provide an old object ID.
Stop unsetting the flag so that we can use it as this described signal
in a subsequent commit. Skip checking the old object ID for log-only
updates so that we don't expect it to match the current on-disk state.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
refs/files: detect race when generating reflog entry for HEAD
When updating a reference that is being pointed to HEAD we don't only
write a reflog message for that particular reference, but also generate
one for HEAD. This logic is handled by `split_head_update()`, where we:
1. Verify that the condition actually triggered. This is done by
reading HEAD at the start of the transaction so that we can then
check whether a given reference update refers to its target.
2. Queue a new log-only update for HEAD in case it did.
But the logic is unfortunately not free of races, as we do not lock the
HEAD reference after we have read its target. This can lead to the
following two scenarios:
- HEAD gets concurrently updated to point to one of the references we
have already processed. This causes us not writing a reflog message
even though we should have done so.
- HEAD gets concurrently updated to no longer point to a reference
anymore that we have already processed. This causes us to write a
reflog message even though we should _not_ have done so.
Improve the situation by introducing a new `REF_LOG_VIA_SPLIT` flag that
is specific to the "files" backend. If set, we will double check that
the HEAD reference still points to the reference that we are creating
the reflog entry for after we have locked HEAD. Furthermore, instead of
manually resolving the old object ID of that entry, we now use the same
old state as for the parent update.
If we detect such a racy update we abort the transaction. This is a bit
heavy-handed: the user didn't even ask us to write a reflog update for
"HEAD", so it might be surprising if we abort the transaction. That
being said:
- Normal users wouldn't typically hit this case as we only hit the
relevant code when committing to a branch that is being pointed to
by "HEAD" directly. Commands like git-commit(1) typically commit to
"HEAD" itself though.
- Scripted users that use git-update-ref(1) and related plumbing
commands are unlikely to hit this case either, as they would have to
update the pointed-to-branch at the same as "HEAD" is being updated,
which is an exceedingly rare event.
The alternative would be to instead drop the log-only update completely,
but that would require more logic that is hard to verify without adding
infrastructure specific for such a test. So we rather do the pragmatic
thing and don't worry too much about an edge case that is very unlikely
to happen.
Unfortunately, this change only helps with the second race. We cannot
reliably plug the first race without locking the HEAD reference at the
start of the transaction. Locking HEAD unconditionally would effectively
serialize all writes though, and that doesn't seem like an option. Also,
double checking its value at the end of the transaction is not an option
either, as its target may have flip-flopped during the transaction.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When migrating reflog entries between different storage formats we must
reconstruct the identity of reflog entries. This is done by passing the
committer passed to the `migrate_one_reflog_entry()` callback function
to `fmt_ident()`.
This results in an invalid identity though: `fmt_ident()` expects the
caller to provide both name and mail of the author, but we pass the full
identity as mail. This leads to an identity like:
pks <Patrick Steinhardt ps@pks.im>
Fix the bug by splitting the identity line first. This allows us to
extract both the name and mail so that we can pass them to `fmt_ident()`
separately.
This commit does not yet add any tests as there is another bug in the
reflog migration that will be fixed in a subsequent commit. Once that
bug is fixed we'll make the reflog verification in t1450 stricter, and
that will catch both this bug here and the other bug.
Note that we also add two new `name` and `mail` string buffers to the
callback structures and splice them through to the callbacks. This is
done so that we can avoid allocating a new buffer every time we compute
the committer information.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The last parameter in `split_ident_line()` is the length of the line
passed in by the caller. As such, most callers pass in either the result
of `strlen()`, `struct strbuf::len` or a pointer diff, all of which
are expected to be positive numbers. Regardless of that, the function
accepts a signed integer, which is somewhat confusing.
Fix the function signature to instead accept a `size_t`.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin/reflog: implement subcommand to write new entries
While we provide a couple of subcommands in git-reflog(1) to remove
reflog entries, we don't provide any to write new entries. Obviously
this is not an operation that really would be needed for many use cases
out there, or otherwise people would have complained that such a command
does not exist yet. But the introduction of the "reftable" backend
changes the picture a bit, as it is now basically impossible to manually
append a reflog entry if one wanted to do so due to the binary format.
Plug this gap by introducing a simple "write" subcommand. For now, all
this command does is to append a single new reflog entry with the given
object IDs and message to the reflog. More specifically, it is not yet
possible to:
- Write multiple reflog entries at once.
- Insert reflog entries at arbitrary indices.
- Specify the date of the reflog entry.
- Insert reflog entries that refer to nonexistent objects.
If required, those features can be added at a future point in time. For
now though, the new command aims to fulfill the most basic use cases
while being as strict as possible when it comes to verifying parameters.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a subsequent commit we'll add another user that wants to write reflog
entries. This requires them to call `ref_transaction_update_reflog()`,
but that function is local to "refs.c".
Export the function to prepare for the change. While at it, drop the
`flags` field, as all callers are for now expected to use the same flags
anyway.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The way subcommands of git-reflog(1) are laid out does not make any
immediate sense. Reorder them such that read-only subcommands precede
writing commands for a bit more structure.
Furthermore, move the "expire" subcommand last. This prepares for a
subsequent change where we are about to introduce a new "write" command
to append reflog entries. Like this, the writing subcommands are ordered
such that those affecting a single reflog come before those spanning
across all reflogs.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-reflog: convert to use synopsis type
With 974cdca345c (doc: introduce a synopsis typesetting, 2024-09-24) we
have introduced a new synopsis type that simplifies the rules for
typesetting a command's synopsis. Convert the git-reflog(1)
documentation to use it.
While at it, convert the list of options to use backticks. This is done
to appease an upcoming new linter that mandates the use of backticks
when using the synopsis type.
Signed-off-by: Patrick Steinhardt <ps@pks.im> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Thu, 24 Jul 2025 23:03:56 +0000 (16:03 -0700)]
Merge branch 'ss/compat-bswap-revamp'
Clean-up compat/bswap.h mess.
* ss/compat-bswap-revamp:
bswap.h: provide a built-in based version of bswap32/64 if possible
bswap.h: remove optimized x86 version of bswap32/64
bswap.h: always overwrite ntohl/ ntohll macros
bswap.h: define GIT_LITTLE_ENDIAN on msvc as little endian
bswap.h: add support for __BYTE_ORDER__
Junio C Hamano [Thu, 24 Jul 2025 23:03:55 +0000 (16:03 -0700)]
Merge branch 'pw/adopt-c99-bool-officially'
Declare weather-balloon we raised for "bool" type 18 months ago a
success and officially allow using the type in our codebase.
* pw/adopt-c99-bool-officially:
strbuf: convert predicates to return bool
git-compat-util: convert string predicates to return bool
CodingGuidelines: allow the use of bool
Junio C Hamano [Wed, 23 Jul 2025 22:45:16 +0000 (15:45 -0700)]
Merge branch 'ps/sane-ctype-workaround'
Our <sane-ctype.h> header file relied on that the system-supplied
<ctype.h> header is not later included, which would override our
macro definitions, but "amazon linux" broke this assumption. Fix
this by preemptively including <ctype.h> near the beginning of
<sane-ctype.h> ourselves.
* ps/sane-ctype-workaround:
sane-ctype: fix compiler error on Amazon Linux 2
Junio C Hamano [Wed, 23 Jul 2025 22:45:15 +0000 (15:45 -0700)]
Merge branch 'ly/changed-paths-traversal'
Lift the limitation to use changed-path filter in "git log" so that
it can be used for a pathspec with multiple literal paths.
* ly/changed-paths-traversal:
bloom: optimize multiple pathspec items in revision
revision: make helper for pathspec to bloom keyvec
bloom: replace struct bloom_key * with struct bloom_keyvec
bloom: rename function operates on bloom_key
bloom: add test helper to return murmur3 hash
Junio C Hamano [Tue, 22 Jul 2025 20:30:21 +0000 (13:30 -0700)]
Merge branch 'master' of https://github.com/j6t/gitk
* 'master' of https://github.com/j6t/gitk: (21 commits)
gitk: remove header of now empty section "General options"
gitk: separate upstream refs when using the sort-by-type option
gitk: make 'sort-refs-by-type' optional and persistent
gitk: sort by ref type on the 'tags and heads' view
gitk: choosefont - remove a stray debugging line
gitk: allow horizontal commit-graph scrolling
gitk: update aqua scrolling for TclTk 8.6 / TIP171
gitk: update x11 scrolling for TclTk 8.6 / TIP 171
gitk: update win32 scrolling for Tk 8.6 / TIP 171
gitk: mousewheel scrolling functions for Tk 8.6
gitk: wheel scrolling multiplier preference
gitk: separate x11 / win32 / aqua Mouse bindings
gitk: remove non-ttk support code
gitk: replace ${NS} with ttk
gitk: always use themed Tk (ttk)
gitk: use $config_variables as list for save/restore
gitk: remove implementations for Tcl/Tk < 8.6
gitk: Make TclTk 8.6 the minimum, allow 8.7
gitk: remove code targeting git <= 1.7.2
gitk: require git >= 2.20
...
Johannes Sixt [Tue, 22 Jul 2025 16:13:31 +0000 (18:13 +0200)]
Merge branch 'mr/sort-refs-by-type'
* mr/sort-refs-by-type:
gitk: separate upstream refs when using the sort-by-type option
gitk: make 'sort-refs-by-type' optional and persistent
gitk: sort by ref type on the 'tags and heads' view
Mark Levedahl [Sat, 5 Apr 2025 12:00:00 +0000 (08:00 -0400)]
git-gui: eliminate _search_exe
git-gui has _search_exe as needed to give the executable suffix
(.exe) on Windows. But, the prior commit eliminated the only user of
this variable. Delete it.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sun, 6 Apr 2025 15:14:35 +0000 (11:14 -0400)]
git-gui: remove procs gitexec and _git_cmd
gitexec looks up and caches the method to execute git subcommands using
the long deprecated dashed form if found in $(git--exec-path). But,
git_read and git_write now use the dashless form, by-passing gitexec.
This leaves two remaining uses of gitexec: one during startup to define
use of an ssh_key helper, and one in the about dialog box. These are
neither performance critical nor likely to be called more than once, so
do not justify an otherwise unused cacheing system.
Let's change those two uses, making gitexec unused. This allows removing
gitexec and _git_cmd.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 5 Apr 2025 03:08:35 +0000 (23:08 -0400)]
git-gui: use dashless 'git cmd' form for read/write
git-gui implements its own approach to locating and running various git
subcommands, bypassing git's capabilities for running git-*. This was
written in 2007: at that time, many git commands were shell-scripts
stored in $(git --exec-path), git's run-command api was not well adapted
to Windows and had serious performance issues when it worked at all, and
running subcommand 'git foo' as 'git-foo' was common and fully supported.
On Windows, git-gui searches $(git --exec-path) for builtin commands,
then attempts to find an interpreter on PATH to run those, invoking
these differently than on other platforms. For instance, the explicit
shebang #!/usr/bin/perl found in a script will be run by the first Perl
interpreter found on $PATH, which might not be at that specific location
so could be different than what git would run.
The various issues leading to the current implemention no longer exist.
Most git commands are now builtins, links to run those are not installed
in $(git --exec-path) by default (the "dashless" form is recommended
instead), and git's run-command api works well everywhere.
So, let's use git to launch its subcommands on all platforms. Do so by
modifying procs git_read and git_write to use the "dashless" form for
invoking git commands, avoiding the search for git-<foo>. This leaves
_git_cmd unused with cleanup in a later patch.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Mon, 21 Jul 2025 15:12:18 +0000 (11:12 -0400)]
git-gui: default to full copy for linked worktrees
git-gui's default clone method is git-clone's default, and this uses
hardlinks rather than copying the objects directory for local
repositories. However, this method explicitly fails if a symlink (or
.gitfile) exists in the path to the objects directory. Thus, the default
clone option fails for worktrees created by git-new-workdir or
git-worktree. git-gui's original do_clone trapped this error for a
symlinked git-new-workdir tree, directly falling back to a full clone,
while the updated git-gui using git-clone does not. (The old do_clone
could not handle gitfile linked worktrees, however).
Let's apply the more friendly fallback to a full clone in both these
cases where git-clone behavior throws an error on the default method.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Fri, 9 Feb 2024 23:07:45 +0000 (18:07 -0500)]
git-gui: use git-clone
git-gui clones a repository by invoking git-plumbing commands, in proc
do_clone, rather than using git-clone. The justification was that the
low-level commands are guaranteed to provide a stable interface, while
the higher level commands such as git-clone may not be stable. This
approach requires git-gui to continually evolve by mirroring new
features in git itself, which has not happened, while the user interface
in git-clone has proven very stable. Also, git-gui does directly call
many other non-plumbing commands in git's repertoire.
do_clone's last significant functionality change was in 2015, and
updates are required for shallow clones, the reftable backend, cloning
from linked worktrees, and perhaps other features and bugs. For
instance, I had reports of git-gui failing to correctly clone
repositories prior to 2015, resulting in essentially the patch given
here. The only significant work was supporting .gitfile linked worktrees
unknown to do_clone, but supported by git-clone, and none regarding the
interface to git-clone itself. That interface is clearly stable enough
to not be a problem.
Supporting new use-cases with this requires exposing new options in the
clone dialog, then passing flags to git-clone. This avoids updating
do_clone to understand those options, reducing the maintenance burdens.
So, teach git-gui to use git-clone. This change is in one patch as
there is no obvious incremental path to migration. The existing dialog /
options / status screen are unchanged, the known user-visible changes
are that cloning from a working directory linked by a gitfile now works,
there is no auto-fallback to a full copy when cloning linked workdirs
and worktrees (meaning git-clone fails unless a full or shared copy is
selected), and messages displayed are from git-clone.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Junio C Hamano [Mon, 21 Jul 2025 16:14:27 +0000 (09:14 -0700)]
Merge branch 'ja/doc-git-log-markup'
Doc mark-up updates.
* ja/doc-git-log-markup:
doc: git-log: convert log config to new doc format
doc: git-log: convert diff options to new doc format
doc: git-log: convert pretty formats to new doc format
doc: git-log: convert pretty options to new doc format
doc: git-log: convert rev list options to new doc format
doc: git-log: convert line range format to new doc format
doc: git-log: convert line range options to new doc format
doc: git-log convert rev-list-description to new doc format
doc: convert git-log to new documentation format
Junio C Hamano [Mon, 21 Jul 2025 16:14:26 +0000 (09:14 -0700)]
Merge branch 'ps/meson-cleanups'
Meson-based build update.
* ps/meson-cleanups:
ci: use Meson's new `--slice` option
meson: update subproject wrappers
meson: fix lookup of shell on MINGW64
meson: clean up unnecessary variables
meson: improve summary of auto-detected features
meson: stop printing 'https' option twice in our summaries
meson: stop discovering native version of Python
Junio C Hamano [Mon, 21 Jul 2025 16:14:26 +0000 (09:14 -0700)]
Merge branch 'jk/remote-avoid-overlapping-names'
"git remote" now detects remote names that overlap with each other
(e.g., remote nickname "outer" and "outer/inner" are used at the
same time), as it will lead to overlapping remote-tracking
branches.
* jk/remote-avoid-overlapping-names:
remote: detect collisions in remote names
Junio C Hamano [Mon, 21 Jul 2025 16:14:26 +0000 (09:14 -0700)]
Merge branch 'tb/midx-avoid-cruft-packs'
"pack-objects" has been taught to avoid pointing into objects in
cruft packs from midx.
* tb/midx-avoid-cruft-packs:
repack: exclude cruft pack(s) from the MIDX where possible
pack-objects: introduce '--stdin-packs=follow'
pack-objects: swap 'show_{object,commit}_pack_hint'
pack-objects: fix typo in 'show_object_pack_hint()'
pack-objects: perform name-hash traversal for unpacked objects
pack-objects: declare 'rev_info' for '--stdin-packs' earlier
pack-objects: factor out handling '--stdin-packs'
pack-objects: limit scope in 'add_object_entry_from_pack()'
pack-objects: use standard option incompatibility functions
Junio C Hamano [Mon, 21 Jul 2025 16:14:25 +0000 (09:14 -0700)]
Merge branch 'bc/use-sha256-by-default-in-3.0'
Prepare to flip the default hash function to SHA-256.
* bc/use-sha256-by-default-in-3.0:
Enable SHA-256 by default in breaking changes mode
help: add a build option for default hash
t5300: choose the built-in hash outside of a repo
t4042: choose the built-in hash outside of a repo
t1007: choose the built-in hash outside of a repo
t: default to compile-time default hash if not set
setup: use the default algorithm to initialize repo format
Use legacy hash for legacy formats
builtin: use default hash when outside a repository
hash: add a constant for the legacy hash algorithm
hash: add a constant for the default hash algorithm
Michael Rappazzo [Sat, 19 Jul 2025 19:24:39 +0000 (15:24 -0400)]
gitk: separate upstream refs when using the sort-by-type option
Since the upstream refs of local refs may be of more significance in the
context of the local refs, they are sorted after local refs and before the
remainder of the remote refs.
Signed-off-by: Michael Rappazzo <michael.rappazzo@infor.com> Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Michael Rappazzo [Fri, 18 Jul 2025 20:33:08 +0000 (16:33 -0400)]
gitk: make 'sort-refs-by-type' optional and persistent
On the 'tags and heads' view, add an option to enable or disable
'Sort refs by type'. This option is read from and written to the
config file. Clicking on the option will update the refs in the
view.
Signed-off-by: Michael Rappazzo <michael.rappazzo@infor.com> Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Mark Levedahl [Mon, 14 Jul 2025 16:15:49 +0000 (12:15 -0400)]
git-gui: remove ${NS} indirection for ttk
git-gui uses ${NS} to switch between non-themed and themed widgets, with
${NS} == 'ttk' selecting the latter. As git-gui now always uses ttk,
this indirection is not needed. Remove it.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Wed, 21 May 2025 20:31:14 +0000 (16:31 -0400)]
git-gui: always use themed widgets from ttk
git-gui optionally uses themed ui elements from ttk, but the full set of
ttk ui elements is always available with Tk 8.6. Keeping code making
ttk use optional increases maintenance burden for no benefit. Let's use
ttk always, allowing removal of alternate code paths in subsequent
patches.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sun, 18 Feb 2024 18:06:05 +0000 (13:06 -0500)]
git-gui: remove redundant check for Tk >= 8.5
Since commit c80d7be5e1e0d, git-gui checks for the availability of ttk
before enabling its use, but this check is redundant as Tk >= 8.6 is
required. Remove the redundant check.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Fri, 16 Feb 2024 23:24:06 +0000 (18:24 -0500)]
git-gui: remove unreachable Tk 8.4 code
git-gui has remnant code to allow some drawing with Tk 8.4 predating the
addition of themed widgets. As git-gui requires Tk >= 8.6, this code can
never trigger. Remove it.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 05:19:56 +0000 (00:19 -0500)]
git-gui: remove unused git-version
git-version supports choosing different bodies of code passed into it,
rather than using the more traditional if/else construct typically used.
The only use of git-version in this mode was by its author in 2007, and
that code has been deleted. So, delete this now unused function that
was mostly ignored.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Fri, 9 Feb 2024 22:58:04 +0000 (17:58 -0500)]
git-gui: use git_init to create new repository dir
When creating a new repository, git-gui creates a directory, cds to it,
then runs git-init, but git-init learned to create and initialize the
directory in 1.6.5. git-gui requires git version >= 2.36, so teach
git-gui to use git-init's full capability.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 05:13:45 +0000 (00:13 -0500)]
git-gui: git-remote is always available
git-gui checks for git version >= 1.6.6 before enabling the remotes
menu. But git-gui requires git v2.36 or later, so git-remote is always
available. Delete this check and always enable the menu.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
git-gui's merge driver includes code to invoke the recursive strategy
for merging prior to git v2.5 that added a simpler syntax. As git-gui
requires git v2.36 or later, let's delete the code targeting earlier
git.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 05:11:32 +0000 (00:11 -0500)]
git-gui: git-diff knows submodules and textconv
git-gui's diff functions avoid using textconv filters on git < 1.6.1, or
asking about submodules on version before 1.7.2, but git-gui requires
git >= v2.36. So, remove this now obsolete code.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 05:09:02 +0000 (00:09 -0500)]
git-gui: git-blame understands -w and textconv
git-gui uses alternate code paths for git versions < 1.7.2, avoiding use
of --ignore-all-space and textconv. git-gui requires git v2.36 or later,
so this alternate code is obsolete. Remove it.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 03:03:30 +0000 (22:03 -0500)]
git-gui: git rev-parse knows show_toplevel
git-gui has its own code to determine the worktree root for git-versions
earlier than 1.7.0, where git rev-parse learned this function. git-gui
requires git v2.36 or later, so delete the now obsolete alternate code.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Mon, 12 Feb 2024 19:42:05 +0000 (14:42 -0500)]
git-gui: use git-branch --show-current
git-gui relies upon the files back-end to determine the current branch.
This does not support the newer reftables backend. But, git-branch has
long supported --show-current to get this same information regardless of
backend cahnged. So teach git-gui to use git-branch --show-current.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 5 Apr 2025 14:19:48 +0000 (10:19 -0400)]
git-gui: git-diff-index always knows submodules
git-gui asks for submodule info only on git-versions >=1.72, which
introduced such capability. But, git-gui requires git version >= 2.36,
so this alternate code path is obsolete. Remove it.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 5 Apr 2025 14:18:06 +0000 (10:18 -0400)]
git-gui: git ls-files knows --exclude-standard
git-gui includes code to implement ls-files for git versions prior to
1.63 that did not know --exclude-standard. But, git-gui now requires git
version >= 2.36, so remove the obsolete code.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 17 May 2025 02:25:17 +0000 (22:25 -0400)]
git-gui: Make TclTk 8.6 the minimum, allow 8.7
git-gui requires that Tcl and Tk are 8.5, though the check using
'package require' allows 8.6. As git-gui runs under wish, both Tcl and
Tk are always available and of the same version, so only one need be
checked.
The 8.5 requirement is very outdated as the earliest Tcl currently
shipping on any supported OS is 8.6. 8.7 is in alpha test and is
generally compatible with 8.6, so should also be allowed. Tcl 9.0 has
planned compatibility breaking changes so cannot be allowed.
Let's update the requirements to be 8.6 or 8.7, and check only on Tcl as
Tk will be the same version.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 13 Feb 2024 04:32:44 +0000 (23:32 -0500)]
git-gui: require git >= 2.36
git-gui since commit d6967022 explicitly requires version >= 1.5.0, and
this coded requirement has never been changed. But, since 0730a5a3a
git-gui actually requires git 2.36, providing 'git hook run.' git-gui
throws an error if that command is not supported.
So, let's update the requirement checking code to 2.36, and throw a more
useful error if this is not met.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 27 May 2025 03:47:39 +0000 (23:47 -0400)]
gitk: allow horizontal commit-graph scrolling
gitk commit 5fdcbb1390 ("gitk: Fixes for Mac OS X TkAqua", 2009-03-23),
adds horizontal scrolling of the commit graph pane on aqua, but not on
x11 or win32. Also, the horizontal scrolling is triggered by MouseWheel
events attached to any of the three panes, not just the commit graph
that is the only one that scrolls. It is unusual to scroll a widget that
is not under the mouse, many would consider this a bug. No horizontal
scrollbar is provided for this, so there is no real cue for the user
that horizontal scrolling is available. We removed this aqua only
feature by transitioning aqua to use the common MouseWheel bindings set.
Let's add this as a feature on all platforms, and use the same approach
for scaling scroll motion as we do elsewhere. For horizontal scrolling,
honor only events received by the commit graph in conformance with
normal GUI design. Vertical scrolling is unchanged, and events received
by any of the 3 panes continue to scroll all 3 in unison.
Per the ancient and long ignored CUA standards, we should add a
horizontal scrollbar to the commit-graph, but gitk's interface is
already very cluttered: adding a scrollbar to only one of these three
panes is difficult while maintaining common pane vertical size,
especially so considering the movable sash separating panes 1 & 2, and
will consume yet more space. So, leave this as a hidden feature, now
available on all platforms.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 3 Jun 2025 19:04:27 +0000 (15:04 -0400)]
gitk: update aqua scrolling for TclTk 8.6 / TIP171
Tk provides MouseWheel events to aqua, similar to win32. But, these
events on aqua have a nominal motion value (%D) of 1, not 120 as on
win32. gitk on aqua provides specific bindings only for the top 3 panes,
giving a nominal scrolling amount of +/- 1 for all events. gitk includes
a hidden feature providing horizontal scrolling of the commit graph,
added in 5fdcbb1390 ("gitk: Fixes for Mac OS X TkAqua", 2009-03-23).
This horizontal scrolling is triggered by mouse events in any of the top
3 panes, and thus violates normal gui design where the object under the
mouse cursor scrolls.
Let's update this using the common bindings in 'proc bind_mousewheel',
allowing user preferences on motion scaling to apply to all windows.
The commit graph scrolling feature is removed by this, and will be added
back for all platforms in a later commit.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Fri, 6 Jun 2025 04:03:33 +0000 (00:03 -0400)]
gitk: update x11 scrolling for TclTk 8.6 / TIP 171
gitk has x11 mouse bindings that receive button presses, not MouseWheel
events, as this is the Tk implementation through Tk 8.6. On x11, gitk
translates each button event to a scrolling value of +/- 5 for the upper
three panes that scroll vertically as one unit. gitk applies similar
scaling for horizontal scaling of the lower-left commit details pane
(ctext), but not for vertical scrolling of either of the bottom panes.
Rather, the Tk default scrolling actions are used for vertical
scrolling.
Let's make X11 behave similarly to the just modified win32 platform. Do
so by connecting vertical and horizontal scrolling events for the same
items bound in 'proc bind_mousewheel' and using the same user preference
values.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 6 Jul 2024 15:08:32 +0000 (11:08 -0400)]
gitk: update win32 scrolling for Tk 8.6 / TIP 171
gitk on win32 binds windows_mousewheel_redirector to all MouseWheel
events in the main window. This proc determines the widget under the
cursor, then determines what scroll command to give, possibly none, and
issues scroll commands to the widget. The top panes get only vertical
scroll events, as does the lower right Patch/Tree pane. All others get
both vertical and horizontal events. These are all hard coded at +/-
five lines.
We now have common MouseWheel event bindings that follow user
preferences for the scrolling amount, bind for only the five main
display widgets, and leave the other gui elements untouched. Let's use
this instead. With the scrolling preference set at 5, the users should
not notice much, if any, difference.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Fri, 6 Jun 2025 01:45:22 +0000 (21:45 -0400)]
gitk: mousewheel scrolling functions for Tk 8.6
gitk supports scrolling of 5 windows, but does this differently on the
aqua, x11, and win32 platforms as Tk provides different events on each.
TIP 171 removes some differences on win32 while altering the required
bindings on x11. TIP 474, which is in Tk 8.7 and later, finally unifies
all platforms on using common MouseWheel bindings. Importantly for now,
TIP 171 causes delivery of MouseWheel events to the widget under the
mouse cursor on win32, eliminating the need for completely different
bindings on win32.
Let's make some common functions to unify as much as we can in Tk 8.6.
Examining the platforms shows that the default platform scrolling is
overridden differently on the 3 platforms, and the nominal amount of
motion achieved per mouse wheel "click" is different. win32 nominally
makes everything move 5 lines per click, aqua 1 line per click, and x11
is a mixture. Part of this is due to win32 overriding all scroll events,
while x11 and aqua override smaller sets. Also, note that the text
widgets (the lower two panes) always scroll by 2-3 lines when given a
smaller scroll amount, while the upper three canvas objects follow the
requested scrolling value more accurately.
First, let's have a common routine to calculate the scroll value to give
to a widget in an event. This accounts for the user preference, the
scale of the %D (delta) value given by the event (120 on win32, 1 on
aqua, assumed 1 on x11), and must always be integer. Include negation as
by convention the screen moves opposite to the MouseWheel delta. Allow
setting an offset value to account for the larger minimum scrolling of
text widgets.
Second, let's have a common declaration of MouseWheel event bindings, as
those are shared by all in Tcl9, and by aqua/win32 earlier. Bind all
five display windows here. Note that the Patch/Tree widget (cflist)
cannot scroll horizontally.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 3 Jun 2025 11:36:44 +0000 (07:36 -0400)]
gitk: wheel scrolling multiplier preference
gitk provides scrolling of several windows, uses hard-coded values for
the amount of scrolling, and these values differ across platforms and
widgets. The nominal value used is either 1 text line per mouse /
touchpad / button event, or 5 lines. Furthermore, Tk does not scroll
text widgets by 1 line when told to, this usually gets 2-3 lines of
motion. The upper canvas objects holding the commit graph do scroll as
defined. But, clearly no value is universally preferred, so let's give
the user some control over this. Provide a single multiplier to be
applied for all scroll bindings, with a value of 3 to mean the default
nominal value of 3 line. This is selected both as a compromise between
the various defaults across platforms, and because it is the smallest
value honored by the two text widgets on the bottom of the screen.
Later commits will connect this variable for actual scrolling events.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sat, 6 Jul 2024 17:12:01 +0000 (13:12 -0400)]
gitk: separate x11 / win32 / aqua Mouse bindings
Tk through 8.6 has different approaches for handling mouse wheel /
touchpad scrolling events on the different platforms, and gitk has
separate code for these. But, some x11 bindings are applied on aqua as
we do not have these in a clean if / then / else tree based upon
platform. Let's split these bindings apart.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sun, 8 Jun 2025 12:45:30 +0000 (08:45 -0400)]
gitk: replace ${NS} with ttk
gitk uses ${NS} to select between the original Tk widgets and the newer
themed widgets in ttk. As gitk uses only themed widgets from ttk::,
this indirection now serves no purpose, so let's switch to explicit use
of ttk:: via global search/replace. More simplification, including
removal of the NS variable, is kept for a later patch to keep this one
smaller.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Sun, 8 Jun 2025 12:16:34 +0000 (08:16 -0400)]
gitk: always use themed Tk (ttk)
gitk added the option to used themed Tk (ttk) in 0cc08ff7dd ("gitk: Add
a user preference to enable/disable use of themed widgets", 2009-09-05).
Using ttk had to be optional as Tk 8.4, then in common use, does not
have ttk. ttk is the default when available, so the ttk code paths are
by now very well tested. gitk also has code paths for the older default
widgets, increasing the maintenance burden. Let's make ttk non-optional
to reduce code complexity in later commits.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Tue, 3 Jun 2025 20:17:15 +0000 (16:17 -0400)]
gitk: use $config_variables as list for save/restore
gitk includes many user defined configuration variables, has all of
these are listed in $config_variables. But this list is not used to
define the variables to be loaded, saved, or restored when cancelling
the configuration dialog, and developers must maintain separate lists of
variables for these purposes. This leads to unnecessary errors and merge
conflicts. Let's replace those separate lists with $config_variables to
make maintenance easier.
While we are on topic, sort the list of names in $config_variables.
This makes it simpler to scan and has fewer chances of conflicts
when new names are introduced.
Helped-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
This patch adds the basic support of SHA256 Git repositories.
Most of changes are idiomatic replacement of the hard-coded hash ID
length, but there are subtle things:
* The hash length is determined on startup, and stored in $hashlength
global variable (either 40 or 64).
* The hard-coded "40" are replaced with $hashlength;
for regexp patterns, the ugly string map is used.
* Some code have the fixed numbers like 39 and 45, and those are
replaced with the $hashlength and the offset correction.
* $nullid and $nullid2 are generated for the hash length.
A caveat is that repository picker dialog is performed before
evaluating the repo type, hence $hashlength isn't set there yet.
So the code dealing with the hard-coded "40" are handled differently;
namely, the regexp range is expanded, and the null id is generated
from the HEAD id length locally.
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Junio C Hamano [Wed, 16 Jul 2025 16:42:28 +0000 (09:42 -0700)]
Merge branch 'ag/doc-send-email'
Documentation updates for "git send-email".
* ag/doc-send-email:
docs: mention possible options for Proton Mail users
docs: add a paragraph explaining the `sendmailCmd` option of sendemail
docs: add an OAuth2.0 credential helper for AOL accounts
docs: add outlookidfix config option to sendemail documentation
docs: link OpenSSL's verify(1) manual page to know about -CAfile and -CApath options
Junio C Hamano [Wed, 16 Jul 2025 16:42:26 +0000 (09:42 -0700)]
Merge branch 'ph/fetch-prune-optim'
"git fetch --prune" used to be O(n^2) expensive when there are many
refs, which has been corrected.
* ph/fetch-prune-optim:
clean up interface for refs_warn_dangling_symrefs
refs: remove old refs_warn_dangling_symref
fetch-prune: optimize dangling-ref reporting
Mark Levedahl [Sun, 13 Jul 2025 20:10:33 +0000 (16:10 -0400)]
gitk: Make TclTk 8.6 the minimum, allow 8.7
gitk runs under wish so naturally has Tcl and Tk available and of the
same version. gitk sets a requirement on Tk version >= 8.4: this is very
outdated, and the earliest Tcl currently shipping on any supported OS is
8.6. As 8.7 is in alpha test and is generally compatible with 8.6, we
should allow 8.7. Tcl 9.0 has planned compatibility breaking changes so
is not yet supported.
Let's change the requirements to 8.6-8.7, but not 9.0. Place this at the
top of file so the requirements are obvious.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Thu, 5 Jun 2025 21:18:19 +0000 (17:18 -0400)]
gitk: remove code targeting git <= 1.7.2
gitk has a few code fragments that are used only for git versions <=
1.7.2 that do not support submodules, notes, word differences, or
textconv filters. We just set the minimum git version higher than 1.7.2
so these code fragments have no effect. Delete them.
Helped-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Mark Levedahl [Thu, 5 Jun 2025 21:09:02 +0000 (17:09 -0400)]
gitk: require git >= 2.20
gitk has alternate code paths for early git up to 1.72, and has no
defined minimum version. Setting any version > 1.72 as minimum will
allow removing those code paths.
The recent set of advisories published for git, gitk, and git-gui add
updates for v2.43 and later, but Debian has buster withgit 2.20 still
available. While Debian would be responsible for backporting any fixes
to such an early version, we have no good reason preclude it.
So, make 2.20 the minimum required git version.
Helped-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Mark Levedahl <mlevedahl@gmail.com>
Now that the string predicates defined in git-compat-util.h all
return bool let's convert the return type of the string predicates
in strbuf.{c,h} to match them.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-compat-util: convert string predicates to return bool
Since 8277dbe987 (git-compat-util: convert skip_{prefix,suffix}{,_mem}
to bool, 2023-12-16) a number of our string predicates have been
returning bool instead of int. Now that we've declared that experiment
a success, let's convert the return type of the case-independent
skip_iprefix() and skip_iprefix_mem() functions to match the return
type of their case-dependent equivalents. Returning bool instead of
int makes it clear that these functions are predicates.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We have had a test balloon for C99's bool type since 8277dbe987
(git-compat-util: convert skip_{prefix,suffix}{,_mem} to bool,
2023-12-16). As we've had it over 18 months without any complaints
let's declare it a success.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Wed, 16 Jul 2025 00:25:23 +0000 (00:25 +0000)]
SubmittingPatches: allow non-real name contributions
Our submission guidelines require people to use their real name, but
this is not always suitable for various reasons.
For people who are transgender or non-binary and are transitioning or
who think they might want to transition, it can be a major obstacle and
cause major discomfort to require the use of their real name. This is
made worse by the fact that Git provides no way to change names built
into history, so the use of a deadname is forever. Our code of conduct
states that we "pledge to act and interact in ways that contribute to an
open, welcoming, diverse, inclusive, and healthy community," and
changing this policy is one way we can improve things for contributors.
In addition, there are some developers who are so widely known
pseudonymously that they have a Wikipedia page with their handle and no
real name. It would seem silly to reject patches from people who are
known and respected in their open-source community just because they
don't wish to share a real name.
There are also other good reasons why people might operate
pseudonymously: because they or their family members are well known and
they wish to protect their privacy, because of current or past
harassment or retaliation or fear of that happening in the future, or
because of concerns about unwanted attention from government officials
or other authority figures. As much as possible, we want to welcome
contributions from anyone who is willing to participate positively in
our community without having them worry about their safety or privacy.
In all of these cases, we should allow people to proceed using a
preferred name or pseudonymously if, in their best judgment, that's the
right thing to do. State that it is common to use a real name but
explicitly mention that contributors who are not comfortable doing so or
prefer to operate pseudonymously or under a preferred name can proceed
otherwise, provided the name is distinctive, identifying, and not
misleading. For instance, using U+2060 (WORD JOINER) as one's ID would
likely be distinctive but not identifying, since most people would have
trouble reading it due to its zero-width nature.
We prohibit identities which are misleading, since our goal is to create
a community which works together with a common goal, and misleading or
deceiving others is not conducive to good community or compatible with
our code of conduct, nor is it compatible with making a legal assertion
about the provenance of one's code.
Explicitly prohibit anonymous contributions to ensure that we have some
line of provenance to a known (if pseudonymous) author who might be able
to respond to questions about it. Explain that this is the reason we
have this policy to help contributors understand the rationale better.
Use "some form of your real name" since some current contributors use
shortened forms of their name or use initials, which have always been
considered acceptable. This helps guide people who would be fine using
their real name but have misconfigured `user.name` thinking it is
intended to be a username or is used for authentication (despite our
documentation to the contrary), but also allows for a variety of
circumstances where the contributor would feel more comfortable not
doing so.
Note that this policy is the same as that of the Linux kernel[0] and the
CNCF[1], as well as many smaller projects. The Linux kernel patch was
Acked-by one of the Linux Foundation's lawyers, Michael Dolan, so it
appears these changes have had legal review.
Additionally, retain the section header ID for ease of linking across
versions.
Ramsay Jones [Tue, 15 Jul 2025 23:32:39 +0000 (00:32 +0100)]
po/meson.build: add missing 'ga' language code
Commit bf5ce434db ("l10n: Add full Irish translation (ga.po)", 2025-05-16)
added a new translation to git. In a make build, new 'po' files (ga.po
in this case) are added to the build automatically using a wildcard
pattern. In a meson build you have to add the language code ('ga') to a
list explicitly to have it included in the build. In order to include the
new translation in the meson build, add the 'ga' language code to the
list of translations in the 'po/meson.build' file.
Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Ramsay Jones [Tue, 15 Jul 2025 23:32:38 +0000 (00:32 +0100)]
meson: fix installation when -Dlibexexdir is set
commit 837f637cf5 ("meson.build: correct setting of GIT_EXEC_PATH",
2025-05-19) corrected the GIT_EXEC_PATH build setting, but then forgot
to update the installation path for the library executables. This causes
a regression when attempting to execute commands, after installing to a
non-standard location (reported here[1]):
$ meson -Dprefix=/tmp/git -Dlibexecdir=libexec-different build
$ meson install
$ /tmp/git/bin/git --exec-path
/tmp/git/libexec-different
$ /tmp/git/bin/git daemon
git: 'daemon' is not a git command. See 'git --help'
In order to fix the issue, use the 'git_exec_path' variable (calculated
while processing -Dlibexecdir) as the 'install_dir' field during the
installation of the library executables.
Junio C Hamano [Tue, 15 Jul 2025 22:18:17 +0000 (15:18 -0700)]
Merge branch 'ps/object-store'
Code clean-up around object access API.
* ps/object-store:
odb: rename `read_object_with_reference()`
odb: rename `pretend_object_file()`
odb: rename `has_object()`
odb: rename `repo_read_object_file()`
odb: rename `oid_object_info()`
odb: trivial refactorings to get rid of `the_repository`
odb: get rid of `the_repository` when handling submodule sources
odb: get rid of `the_repository` when handling the primary source
odb: get rid of `the_repository` in `for_each()` functions
odb: get rid of `the_repository` when handling alternates
odb: get rid of `the_repository` in `odb_mkstemp()`
odb: get rid of `the_repository` in `assert_oid_type()`
odb: get rid of `the_repository` in `find_odb()`
odb: introduce parent pointers
object-store: rename files to "odb.{c,h}"
object-store: rename `object_directory` to `odb_source`
object-store: rename `raw_object_store` to `object_database`
bswap.h: provide a built-in based version of bswap32/64 if possible
The compiler is in general able to recognize the endian shift and
replace it with an optimized opcode if possible. On certain
architectures such as RiscV or MIPS the situation can get complicated.
They don't provide an optimized opcode and masking the "higher" bits may
required loading a constant which needs shifting. This causes the
compiler to emit a lot of instructions for the operation.
The provided builtin directive on these architecture calls a function
which does the operation instead of emitting the code for operation.
Bring back the change from commit 6547d1c9 (bswap.h: add support for
built-in bswap functions, 2025-04-23). The bswap32/64 macro can now be
defined unconditionally so it won't regress on big endian architectures.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc> Signed-off-by: Junio C Hamano <gitster@pobox.com>
bswap.h: remove optimized x86 version of bswap32/64
On x86 the bswap32/64 macro is implemented based on the x86 opcode which
performs the required shifting in just one opcode.
The other CPUs fallback to the generic shifting as implemented by
default_swab32() and default_bswap64() if needed.
I've been looking at how good a compiler is at recognizing the default
shift and emitting an optimized operation:
- x86, arm64 msvc v19.20
default_swab32() optimized
default_bswap64() shifts
_byteswap_uint64() optimized
projects / thirdparty / git.git / log
