The previous two changes introduced a commit walking heuristic for finding
the most likely base branch for a given source. This algorithm walks
first-parent histories until reaching a collision.
This walk _should_ be very fast. Exceptions include cases where a
commit-graph file does not exist, leading to a full walk of all reachable
commits to compute generation numbers, or a case where no collision in the
first-parent history exists, leading to a walk of all first-parent history
to the root commits.
The p1500 test script guarantees a complete commit-graph file during its
setup, so we will not test that scenario. Do create a new root commit in an
effort to test the scenario of parallel first-parent histories.
Even with the extra root commit, these tests take no longer than 0.02
seconds on my machine for the Git repository. However, the results are
slightly more interesting in a copy of the Linux kernel repository:
Test
---------------------------------------------------------------
1500.2: ahead-behind counts: git for-each-ref 0.12
1500.3: ahead-behind counts: git branch 0.12
1500.4: ahead-behind counts: git tag 0.12
1500.5: contains: git for-each-ref --merged 0.04
1500.6: contains: git branch --merged 0.04
1500.7: contains: git tag --merged 0.04
1500.8: is-base check: test-tool reach (refs) 0.03
1500.9: is-base check: test-tool reach (tags) 0.03
1500.10: is-base check: git for-each-ref 0.03
1500.11: is-base check: git for-each-ref (disjoint-base) 0.07
Signed-off-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous change introduced the get_branch_base_for_tip() method in
commit-reach.c. The motivation of that change was about using a heuristic to
deteremine the base branch for a source commit from a list of candidate
commit tips. This change makes that algorithm visible to users via a new
atom in the 'git for-each-ref' format. This change is very similar to the
chang in 49abcd21da (for-each-ref: add ahead-behind format atom,
2023-03-20).
Introduce the 'is-base:<source>' atom, which will indicate that the
algorithm should be computed and the result of the algorithm is reported
using an indicator of the form '(<source>)'. For example, using
'%(is-base:HEAD)' would result in one line having the token '(HEAD)'.
Use the sorted order of refs included in the ref filter to break ties in the
algorithm's heuristic. In the previous change, the motivating examples
include using an L0 trunk, long-lived L1 branches, and temporary release
branches. A caller could communicate the ordered preference among these
categories using the input refpecs and avoiding a different sort mechanism.
This sorting behavior is tested in the test scripts.
It is important to include this atom as a special case to
can_do_iterative_format() to match the expectations created in bd98f9774e
(ref-filter.c: filter & format refs in the same callback, 2023-11-14). The
ahead-behind atom was one of the special cases, and this similarly requires
using an algorithm across all input refs before starting the format of any
single ref.
In the test script, the format tokens use colons or lack whitespace to avoid
Git complaining about trailing whitespace errors.
Signed-off-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The lookup_commit_reference_by_name() method uses lookup_commit_reference()
without an option to use lookup_commit_reference_gently(). Create a gentle
version of the method so it can be used in locations where non-commits may
be found but error messages should be silenced.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a new reachability algorithm that intends to discover (from a heuristic)
which branch was used as the starting point for a given commit. Add focused
tests using the 'test-tool reach' command.
In repositories that use pull requests (or merge requests) to advance one or
more "protected" branches, the history of that reference can be recovered by
following the first-parent history in most cases. Most are completed using
no-fast-forward merges, though squash merges are quite common. Less common
is rebase-and-merge, which still validates this assumption. Finally, the
case that breaks this assumption is the fast-forward update (with potential
rebasing). Even in this case, the previous commit commonly appears in the
first-parent history of the branch.
Similar assumptions can be made for a topic branch created by a single user
with the intention to merge back into another branch. Using 'git commit',
'git merge', and 'git cherry-pick' from HEAD will default to having the
first-parent commit be the previous commit at HEAD. This history changes
only with commands such as 'git reset' or 'git rebase', where the command
names also imply that the branch is starting from a new location.
With this movement of branches in mind, the following heuristic is proposed
as a way to determine the base branch for a given source branch:
Among a list of candidate base branches, select the candidate that
minimizes the number of commits in the first-parent history of the source
that are not in the first-parent history of the candidate.
Prior third-party solutions to this problem have used this optimization
criteria, but have relied upon extracting the first-parent history and
comparing those lists as tables instead of using commit-graph walks.
Given current command-line interface options, this optimization criteria is
not easy to detect directly. Even using the command
git rev-list --count --first-parent <base>..<source>
does not measure this count, as it uses full reachability from <base> to
determine which commits to remove from the range '<base>..<source>'. This
may lead to one asking if we should instead be using the full reachability
of the candidate and only the first-parent history of the source. This,
unfortunately, does not work for repositories that use long-lived branches
and automation to merge across those branches.
In extremely large repositories, merging into a single trunk may not be
feasible. This is usually due to the desired frequency of updates
(thousands of engineers doing daily work) combined with the time required to
perform a validation build. These factors combine to create significant
risk of semantic merge conflicts, leading to build breaks on the trunk. In
response, repository maintainers can create a single Level Zero (L0) trunk
and multiple Level One (L1) branches. By partitioning the engineers by
organization, these engineers may see lower risk of semantic merge conflicts
as well as be protected against build breaks in other L1 branches. The key
to making this system work is a semi-automated process of merging L1
branches into the L0 trunk and vice-versa. In a large enough organization,
these L1 branches may further split into L2 or L3 branches, but the same
principles apply for merging across deeper levels.
If these automated merges use a typical merge with the second parent
bringing in the "new" content, then each L0 and L1 branch can track its
previous positions by following first-parent history, which appear as
parallel paths (until reaching the first place where the branches diverged).
If we also walk to second parents, then the histories overlap significantly
and cannot be distinguished except for very-recent changes.
For this reason, the first-parent condition should be symmetrical across the
base and source branches.
Another common case for desiring the result of this optimization method is
the use of release branches. When releasing a version of a repository, a
branch can be used to track that release. Any updates that are worth fixing
in that release can be merged to the release branch and shipped with only
the necessary fixes without any new features introduced in the trunk branch.
The 'maint-2.<X>' branches represent this pattern in the Git project. The
microsoft/git fork uses 'vfs-2.<X>.<Y>' branches to track the changes that
are custom to that fork on top of each upstream Git release 2.<X>.<Y>. This
application doesn't need the symmetrical first-parent condition, but the use
of first-parent histories does not change the results for these branches.
To determine the base branch from a list of candidates, create a new method
in commit-reach.c that performs a single* commit-graph walk. The core
concept is to walk first-parents starting at the candidate bases and the
source, tracking the "best" base to reach a given commit. Use generation
numbers to ensure that a commit is walked at most once and all children have
been explored before visiting it. When reaching a commit that is reachable
from both a base and the source, we will then have a guarantee that this is
the closest intersection of first-parent histories. Track the best base to
reach that commit and return it as a result. In rare cases involving
multiple root commits, the first-parent history of the source may never
intersect any of the candidates and thus a null result is returned.
* There are up to two walks, since we require all commits to have a computed
generation number in order to avoid incorrect results. This is similar to
the need for computed generation numbers in ahead_behind() as implemented
in fd67d149bd (commit-reach: implement ahead_behind() logic, 2023-03-20).
In order to track the "best" base, use a new commit slab that stores an
integer. This value defaults to zero upon initialization, so use -1 to
track that the source commit can reach this commit and use 'i + 1' to track
that the ith base can reach this commit. When multiple bases can reach a
commit, minimize the index to break ties. This allows the caller to specify
an order to the bases that determines some amount of preference when the
heuristic does not result in a unique result.
The trickiest part of the integer slab is what happens when reaching a
collision among the histories of the bases and the history of the source.
This is noticed when viewing the first parent and seeing that it has a slab
value that differs in sign (negative or positive). In this case, the
collision commit is stored in the method variable 'branch_point' and its
slab value is set to -1. The index of the best base (so far) is stored in
the method variable 'best_index'. It is possible that there are multiple
commits that have the branch_point as its first parent, leading to multiple
updates of best_index. The result is determined when 'branch_point' is
visited in the commit walk, giving the guarantee that all commits that could
reach 'branch_point' were visited.
Several interesting cases of collisions and different results are tested in
the t6600-test-reach.sh script. Recall that this script also tests the
algorithm in three possible states involving the commit-graph file and how
many commits are written in the file. This provides some coverage of the
need (and lack of need) for the ensure_generations_valid() method.
Signed-off-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Work around asciidoctor's css that renders `monospace` material
in the SYNOPSIS section of manual pages as block elements.
* js/doc-markup-updates-fix:
Doc: fix Asciidoctor css workaround
asciidoctor: fix `synopsis` rendering
Repacking a repository with multi-pack index started making stupid
pack selections in Git 2.45, which has been corrected.
* ds/midx-write-repack-fix:
midx-write: revert use of --stdin-packs
t5319: add failing test case for repack/expire
The previous step introduced docinfo.html to be used to tweak the
CSS used by the asciidoctor, that by default renders <code> inside
<pre> as a block element, breaking the SYNOPSIS section of a few
pages that adopted a new convention we use since Git 2.45.
But in this project, HTML files are all generated. We do not force
any human to write HTML by hand, which is an unusual and cruel
punishment. "*.html" is in the .gitignore file, and "make clean"
removes them. Having a tracked .html file makes "make clean" make
the tree dirty by removing the tracked docinfo.html file.
Let's do an obvious, minimum and stupid workaround to generate that
file at runtime instead. The mark-up is being rethought in a major
way for the next development cycle, and the CSS workaround we added
in the previous step may have to adjusted, possibly in a large way,
anyway.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It was reported that t1460-refs-migrate.sh fails when using Cygwin with
errors like the following:
error: could not link file '.git/ref_migration.sr9pEF/reftable' to '.git/reftable': Permission denied
As some debugging surfaced, the root cause of this is that some files of
the newly-initialized ref store are still open when the target format is
the "reftable" format, and Cygwin refuses to rename open files.
Fix this issue by closing the new ref store before renaming its files
into place. This is a slight change in behaviour compared to before,
where we kept the new ref store open and then updated the repository's
ref store to point to it.
While we could re-open the new ref store after we have moved files
around, this is ultimately unnecessary. We know that the only user of
`repo_migrate_ref_storage_format()` is the git-refs(1) command, and it
won't access the ref store after it has been migrated anyway. So
reinitializing the ref store would be a waste of time. Regardless of
that it is still sensible to leave the repository in a consistent state.
But instead of reinitializing the ref store, we can simply unset the
repo's ref store altogether and let `get_main_ref_store()` lazily
initialize the new ref store as required.
Reported-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 76880f0510 (doc: git-clone: apply new documentation formatting
guidelines, 2024-03-29), the synopsis of `git clone`'s manual page is
rendered differently than before; Its parent commit did the same for
`git init`.
The result looks quite nice. When rendered with AsciiDoc, that is. When
rendered using AsciiDoctor and displayed in a graphical web browser such
as Firefox, Chrome, Edge, etc, the result is quite unpleasant to my eye,
reading something like this:
SYNOPSIS
git clone
[
--template=
<template-directory>]
[
-l
] [
-s
] [
--no-hardlinks
] [
-q
] [
[... continuing like this ...]
The reason is that AsciiDoctor's default style sheet contains this (see
https://github.com/asciidoctor/asciidoctor/blob/854923b15533/src/stylesheets/asciidoctor.css#L519-L521
for context):
pre > code {
display: block;
}
It is this `display: block` that forces the parts that are enclosed in
`<code>` tags (such as the `git clone` or the `--template=` part) to be
rendered on their own line.
Side note: This seems not to affect console web browsers like `lynx` or
`w3m`, most likely because most style sheet directions cannot be
respected in text terminals and therefore they seem to punt on style
sheets altogether.
To fix this, let's apply the method recommended by AsciiDoctor in
https://docs.asciidoctor.org/asciidoctor/latest/html-backend/default-stylesheet/#customize-docinfo
to partially override AsciiDoctor's default style sheet so that the
`<code>` sections of the synopsis are no longer each rendered on their
own, individual lines.
This fixes https://github.com/git-for-windows/git/issues/5063.
Even on the Git home page, where AsciiDoctor's default stylesheet is
_not_ used, this change resulted in some unpleasant rendering where not
only the font is changed for the `<code>` sections of the synopsis, but
padding and a different background color make the visual impression
quite uneven. This has been addressed in the meantime, via
https://github.com/git/git-scm.com/commit/a492d0565512.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Asciidoc.py does not have the concept of generalized roles, whereas
asciidoctor interprets [foo]`blah` as blah with role foo in the
synopsis, making in effect foo disappear in the output. Note that
square brackets not directly followed by an inline markup do not
define a role, which is why we do not have the issue on other parts of
the documentation.
In order to get a consistant result across asciidoctor and
asciidoc.py, the hack is to use the {empty} entity
to split the bracket part from the inline format part.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This reverts b7d6f23a17 (midx-write.c: use `--stdin-packs` when
repacking, 2024-04-01) and then marks the test created in the previous
change as passing.
The fundamental issue with the reverted change is that the focus on
pack-files separates the object selection from how the multi-pack-index
selects a single pack-file for an object ID with multiple copies among
the tracked pack-files.
The change was made with the intention of improving delta compression in
the resulting pack-file, but that can be resolved with the existing
object list mechanism. There are other potential pitfalls of doing an
object walk at this time if the repository is a blobless partial clone,
and that will require additional testing on top of the one that changes
here.
Signed-off-by: Derrick Stolee <stolee@gmail.com>
Acked-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Git 2.45.0 included the change b7d6f23a17 (midx-write.c: use
`--stdin-packs` when repacking, 2024-04-01) which caused the 'git
multi-pack-index repack' command to use 'git pack-objects --stdin-packs'
instead of listing the objects to repack. While this change was
motivated by efficient cross-process communication and the ability to
improve delta compression, it breaks a fundamental function of the
'incremental-repack' task that is enabled by default in Scalar clones or
Git repositories that run 'git maintenance start'.
The 'incremental-repack' task performs a two-step process of the
'expire' and 'repack' subcommands of the 'git multi-pack-index' builtin.
The 'expire' command removes any pack-files listed in the
multi-pack-index but without any referenced objects. The 'repack' task
then finds a batch of pack-files to repack and sends their objects to
'git pack-objects'. Both the pack-files chosen for the batch and the
objects chosen to repack are based on the ones that the multi-pack-index
references. Objects that appear in a pack-file but have a duplicate copy
in a newer pack-file are not considered in this case. Since the
multi-pack-index references only the newest copy of an object, this
allows the next 'incremental-repack' task to remove the pack-files in
the next 'expire' task. This delay is intentional due to how Windows
handles may block deletion of files with open read handles.
However, the mentioned commit changed this behavior to divorce the set
of objects referenced by the multi-pack-index and instead use a set of
"included" and "excluded" pack-files in the 'git pack-objects' builtin.
When a pack-file is selected as "included", only the objects it contains
but are not in any "excluded" pack-files are considered for repacking.
This has led to client repositories failing to remove old pack-files as
they still have some referenced objects. This grows over time until the
point that Git is trying to repack the same pack-files over and over.
For now, create a test case that demonstrates the expected behavior, but
also fails in its final line. The setup here it attempting to recreate a
typical situation for a repository that uses a blobless partial clone.
There would be a large initial pack-file from the clone that is never
selected in the 'repack' batch. There are other pack-files that have a
combination of new objects from incremental fetches and possibly blobs
that are not connected to those incremental fetches; these blobs could
be filled in from commands like 'git checkout' or 'git blame'. The
pack-files also have some overlap on purpose so test-1 has some
duplicates in test-2 and test-2 has some duplicates in test-3.
At the end of the test, the test-2 pack-file still exists though it
should have been expired. This test will pass when reverting the
offending commit.
Signed-off-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git var GIT_SHELL_PATH" should report the path to the shell used
to spawn external commands, but it didn't do so on Windows, which
has been corrected.
* js/var-git-shell-path:
var(win32): do report the GIT_SHELL_PATH that is actually used
run-command: declare the `git_shell_path()` function globally
run-command(win32): resolve the path to the Unix shell early
mingw(is_msys2_sh): handle forward slashes in the `sh.exe` path, too
win32: override `fspathcmp()` with a directory separator-aware version
strvec: declare the `strvec_push_nodup()` function globally
run-command: refactor getting the Unix shell path into its own function
What happens when http.cookieFile gets the special value "" has
been clarified in the documentation.
* ps/doc-http-empty-cookiefile:
doc: update http.cookieFile with in-memory cookie processing
"git push '' HEAD:there" used to hit a BUG(); it has been corrected
to die with "fatal: bad repository ''".
* kn/push-empty-fix:
builtin/push: call set_refspecs after validating remote
The http.cookieFile and http.saveCookies configuration variables
have a few values that need to be avoided, which are now ignored
with warning messages.
* jc/http-cookiefile:
http.c: cookie file tightening
The test framework learned to take the test body not as a single
string but as a here-document.
* jk/test-body-in-here-doc:
t/.gitattributes: ignore whitespace in chainlint expect files
t: convert some here-doc test bodies
test-lib: allow test snippets as here-docs
chainlint.pl: add tests for test body in heredoc
chainlint.pl: recognize test bodies defined via heredoc
chainlint.pl: check line numbers in expected output
chainlint.pl: force CRLF conversion when opening input files
chainlint.pl: do not spawn more threads than we have scripts
chainlint.pl: only start threads if jobs > 1
chainlint.pl: add test_expect_success call to test snippets
Tests that use GIT_TEST_SANITIZE_LEAK_LOG feature got their exit
status inverted, which has been corrected.
* rj/test-sanitize-leak-log-fix:
test-lib: GIT_TEST_SANITIZE_LEAK_LOG enabled by default
test-lib: fix GIT_TEST_SANITIZE_LEAK_LOG
When `core.maxTreeDepth` was originally introduced via be20128bfa (add
core.maxTreeDepth config, 2023-08-31), its default value was 4096.
There have since been a couple of updates to its default value that were
not reflected in the documentation for `core.maxTreeDepth`:
- 4d5693ba05 (lower core.maxTreeDepth default to 2048, 2023-08-31)
- b64d78ad02 (max_tree_depth: lower it for MSVC to avoid stack
overflows, 2023-11-01)
Commit 4d5693ba05 lowers the default to 2048 for platforms with smaller
stack sizes, and commit b64d78ad02 lowers the default even further when
Git is compiled with MSVC.
Neither of these changes were reflected in the documentation, which I
noticed while merging newer releases back into GitHub's private fork
(which contained the original implementation of `core.maxTreeDepth`).
Update the documentation to reflect what the platform-specific default
values are.
Noticed-by: Keith W. Campbell <keithc@ca.ibm.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee