The split-index mode had a few corner case bugs fixed.
* tg/split-index-fixes:
travis: run tests with GIT_TEST_SPLIT_INDEX
split-index: don't write cache tree with null oid entries
read-cache: fix reading the shared index for other repos
The http tracing code, often used to debug connection issues,
learned to redact potentially sensitive information from its output
so that it can be more safely sharable.
* jt/http-redact-cookies:
http: support omitting data from traces
http: support cookie redaction when tracing
Merge branch 'sg/travis-build-during-script-phase' into maint
Build the executable in 'script' phase in Travis CI integration, to
follow the established practice, rather than during 'before_script'
phase. This allows the CI categorize the failures better ('failed'
is project's fault, 'errored' is build environment's).
* sg/travis-build-during-script-phase:
travis-ci: build Git during the 'script' phase
parse-options: do not show usage upon invalid option value
Usually, the usage should be shown only if the user does not know what
options are available. If the user specifies an invalid value, the user
is already aware of the available options. In this case, there is no
point in displaying the usage anymore.
This patch applies to "git tag --contains", "git branch --contains",
"git branch --points-at", "git for-each-ref --contains" and many more.
Signed-off-by: Paul-Sebastian Ungureanu <ungureanupaulsebastian@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When computing abbreviation lengths for an object ID against a single
packfile, the method find_abbrev_len_for_pack() currently implements
binary search. This is one of several implementations. One issue with
this implementation is that it ignores the fanout table in the pack-
index.
Translate this binary search to use the existing bsearch_pack() method
that correctly uses a fanout table.
Due to the use of the fanout table, the abbreviation computation is
slightly faster than before. For a fully-repacked copy of the Linux
repo, the following 'git log' commands improved:
The method bsearch_hash() generalizes binary searches using a
fanout table. The only consumer is currently find_pack_entry_one().
It requires a bit of pointer arithmetic to align the fanout table
and the lookup table depending on the pack-index version.
Extract the pack-index pointer arithmetic to a new method,
bsearch_pack(), so this can be re-used in other code paths.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
completion: clear cached --options when sourcing the completion script
The established way to update the completion script in an already
running shell is to simply source it again: this brings in any new
--options and features, and clears caching variables. E.g. it clears
the variables caching the list of (all|porcelain) git commands, so
when they are later lazy-initialized again, then they will list and
cache any newly installed commmands as well.
Unfortunately, since d401f3debc (git-completion.bash: introduce
__gitcomp_builtin, 2018-02-09) and subsequent patches this doesn't
work for a lot of git commands' options. To eliminate a lot of
hard-to-maintain hard-coded lists of options, those commits changed
the completion script to use a bunch of programmatically created and
lazy-initialized variables to cache the options of those builtin
porcelain commands that use parse-options. These variables are not
cleared upon sourcing the completion script, therefore they continue
caching the old lists of options, even when some commands recently
learned new options or when deprecated options were removed.
Always 'unset' these variables caching the options of builtin commands
when sourcing the completion script.
Redirect 'unset's stderr to /dev/null, because ZSH's 'unset' complains
if it's invoked without any arguments, i.e. no variables caching
builtin's options are set. This can happen, if someone were to source
the completion script twice without completing any --options in
between. Bash stays silent in this case.
Add tests to ensure that these variables are indeed cleared when the
completion script is sourced; not just the variables caching options,
but all other caching variables, i.e. the variables caching commands,
porcelain commands and merge strategies as well.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 833622a945 ("stash push: avoid printing errors", 2018-03-19) we
don't use the 'git clean' call for the pathspec case anymore. The
commit however forgot to remove the pathspec argument to the call.
Remove the superfluos argument to make the code a little more obvious.
Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'tg/stash-untracked-with-pathspec-fix' into next
"git stash push -u -- <pathspec>" gave an unnecessary and confusing
error message when there was no tracked files that match the
<pathspec>, which has been fixed.
The transfer.fsckobjects configuration tells "git fetch" to
validate the data and connected-ness of objects in the received
pack; the code to perform this check has been taught about the
narrow clone's convention that missing objects that are reachable
from objects in a pack that came from a promissor remote is OK.
* jt/transfer-fsck-with-promissor:
fetch-pack: do not check links for partial fetch
index-pack: support checking objects but not links
In a way similar to how "git tag" learned to honor the pager
setting only in the list mode, "git config" learned to ignore the
pager setting when it is used for setting values (i.e. when the
purpose of the operation is not to "show").
* ma/config-page-only-in-list-mode:
config: change default of `pager.config` to "on"
config: respect `pager.config` in list/get-mode only
t7006: add tests for how git config paginates
This patch fixes a corner case where diff-highlight may
scramble some diffs when combined with --graph.
Commit 7e4ffb4c17 (diff-highlight: add support for --graph
output, 2016-08-29) taught diff-highlight to skip past the
graph characters at the start of each line with this regex:
($COLOR?\|$COLOR?\s+)*
I.e., any series of pipes separated by and followed by
arbitrary whitespace. We need to match more than just a
single space because the commit in question may be indented
to accommodate other parts of the graph drawing. E.g.:
the diff lines have more spaces between the pipes and the
start of the diff.
However, when we soak up all of those spaces with the
$GRAPH regex, we may accidentally include the leading space
for a context line. That means we may consider the actual
contents of a context line as part of the diff syntax. In
other words, something like this:
normal context line
-old line
+new line
-this is a context line with a leading dash
would cause us to see that final context line as a removal
line, and we'd end up showing the hunk in the wrong order:
normal context line
-old line
-this is a context line with a leading dash
+new line
Instead, let's a be a little more clever about parsing the
graph. We'll look for the actual "*" line that marks the
start of a commit, and record the indentation we see there.
Then we can skip past that indentation when checking whether
the line is a hunk header, removal, addition, etc.
There is one tricky thing: the indentation in bytes may be
different for various lines of the graph due to coloring.
E.g., the "*" on a commit line is generally shown without
color, but on the actual diff lines, it will be replaced
with a colorized "|" character, adding several bytes. We
work around this here by counting "visible" bytes. This is
unfortunately a bit more expensive, making us about twice as
slow to handle --graph output. But since this is meant to be
used interactively anyway, it's tolerably fast (and the
non-graph case is unaffected).
One alternative would be to search for hunk header lines and
use their indentation (since they'd have the same colors as
the diff lines which follow). But that just opens up
different corner cases. If we see:
| | @@ 1,2 1,3 @@
we cannot know if this is a real diff that has been
indented due to the graph, or if it's a context line that
happens to look like a diff header. We can only be sure of
the indent on the "*" lines, since we know those don't
contain arbitrary data (technically the user could include a
bunch of extra indentation via --format, but that's rare
enough to disregard).
Reported-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current flush() helper only shows the queued diff but
does not clear the queue. This is conceptually a bug, but it
works because we only call it once at the end of the
program.
Let's teach it to clear the queue, which will let us use it
in more places (one for now, but more in future patches).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our tests send git's output directly to files or pipes, so
there will never be any color. Let's do at least one --color
test to make sure that we can handle this case (which we
currently can, but will be an easy thing to mess up when we
touch the graph code in a future patch).
We'll just cover the --graph case, since this is much more
complex than the earlier cases (i.e., if it manages to
highlight, then the non-graph case definitely would).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff-highlight: test interleaved parallel lines of history
The graph test in t9400 covers the case of two simultaneous
branches, but all of the commits during this time are on the
right-hand branch. So we test a graph structure like:
| |
| * commit ...
| |
but we never see the reverse, a commit on the left-hand
branch:
| |
* | commit ...
| |
Since this is an easy thing to get wrong when touching the
graph-matching code, let's cover it by adding one more
commit with its timestamp interleaved with the other branch.
Note that we need to pass --date-order to convince Git to
show it this way (since --topo-order tries to keep lines of
history separate).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We generate a bunch of one-line files whose contents match
their names, and then generate our commits by cat-ing those
files. Let's just echo the contents directly, which saves
some processes.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The exact ordering output by Git may depend on the commit
timestamps, so let's make sure they're actually
monotonically increasing, and not all the same (or worse,
subject to how long the test script takes to run).
Let's use test_tick to make sure this is stable. Note that
we actually have to rearrange the order of the branches to
match the expected graph structure (which means that
previously we might racily have been testing a slightly
different output, though the test is written in such a way
that we'd still pass).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The way "git worktree prune" worked internally has been simplified,
by assuming how "git worktree move" moves an existing worktree to a
different place.
* nd/worktree-prune:
worktree prune: improve prune logic when worktree is moved
worktree: delete dead code
gc.txt: more details about what gc does
Conversion from uchar[20] to struct object_id continues.
* bc/object-id: (36 commits)
convert: convert to struct object_id
sha1_file: introduce a constant for max header length
Convert lookup_replace_object to struct object_id
sha1_file: convert read_sha1_file to struct object_id
sha1_file: convert read_object_with_reference to object_id
tree-walk: convert tree entry functions to object_id
streaming: convert istream internals to struct object_id
tree-walk: convert get_tree_entry_follow_symlinks internals to object_id
builtin/notes: convert static functions to object_id
builtin/fmt-merge-msg: convert remaining code to object_id
sha1_file: convert sha1_object_info* to object_id
Convert remaining callers of sha1_object_info_extended to object_id
packfile: convert unpack_entry to struct object_id
sha1_file: convert retry_bad_packed_offset to struct object_id
sha1_file: convert assert_sha1_type to object_id
builtin/mktree: convert to struct object_id
streaming: convert open_istream to use struct object_id
sha1_file: convert check_sha1_signature to struct object_id
sha1_file: convert read_loose_object to use struct object_id
builtin/index-pack: convert struct ref_delta_entry to object_id
...
The gitattributes documentation claims that the pattern
rules are largely the same as for gitignore. However, the
rules for recursion are different.
In an ideal world, we would make them the same (if for
nothing else than consistency and simplicity), but that
would create backwards compatibility issues. For some
discussion, see this thread:
The function ce_write_entry() uses a 'self-initialised' variable
construct, for the symbol 'saved_namelen', to suppress a gcc
'-Wmaybe-uninitialized' warning, given that the warning is a false
positive.
For the purposes of this discussion, the ce_write_entry() function has
three code blocks of interest, that look like so:
The warning implies that gcc thinks it is possible that the first
block is not entered, the calls to copy_cache_entry_to_ondisk()
could toggle the CE_STRIP_NAME flag on, thereby entering block #3
with saved_namelen unset. However, the copy_cache_entry_to_ondisk()
function does not write to ce->ce_flags (it only reads). gcc could
easily determine this, since that function is local to this file,
but it obviously doesn't.
In order to suppress this warning, we make it clear to the reader
(human and compiler), that block #3 will only be entered when the
first block has been entered, by introducing a new 'stripped_name'
boolean variable. We also take the opportunity to change the type
of 'saved_namelen' to 'unsigned int' to match ce->ce_namelen.
Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
-Wuninitialized: remove some 'init-self' workarounds
The 'self-initialised' variables construct (ie <type> var = var;) has
been used to silence gcc '-W[maybe-]uninitialized' warnings. This has,
unfortunately, caused MSVC to issue 'uninitialized variable' warnings.
Also, using clang static analysis causes complaints about an 'Assigned
value is garbage or undefined'.
There are six such constructs in the current codebase. Only one of the
six causes gcc to issue a '-Wmaybe-uninitialized' warning (which will
be addressed elsewhere). The remaining five 'init-self' gcc workarounds
are noted below, along with the commit which introduced them:
1. builtin/rev-list.c: 'reaches' and 'all', see commit 457f08a030
("git-rev-list: add --bisect-vars option.", 2007-03-21).
2. merge-recursive.c:2064 'mrtree', see commit f120ae2a8e ("merge-
recursive.c: mrtree in merge() is not used before set", 2007-10-29).
3. fast-import.c:3023 'oe', see commit 85c62395b1 ("fast-import: let
importers retrieve blobs", 2010-11-28).
4. fast-import.c:3006 'oe', see commit 28c7b1f7b7 ("fast-import: add a
get-mark command", 2015-07-01).
Remove the 'self-initialised' variable constructs noted above.
Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When introducing the stash push feature, and thus allowing users to pass
in a pathspec to limit the files that would get stashed in df6bba0937 ("stash: teach 'push' (and 'create_stash') to honor
pathspec", 2017-02-28), this developer missed one place where the
pathspec should be passed in.
Namely in the call to the 'untracked_files()' function in the
'no_changes()' function. This resulted in 'git stash push -u --
<non-existant>' creating an empty stash when there are untracked files
in the repository other that don't match the pathspec.
As 'git stash' never creates empty stashes, this behaviour is wrong and
confusing for users. Instead it should just show a message "No local
changes to save", and not create a stash.
Luckily the 'untracked_files()' function already correctly respects
pathspecs that are passed to it, so the fix is simply to pass the
pathspec along to the function.
Reported-by: Marc Strapetz <marc.strapetz@syntevo.com> Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
'git stash push -u -- <pathspec>' prints the following errors if
<pathspec> only matches untracked files:
fatal: pathspec 'untracked' did not match any files
error: unrecognized input
This is because we first clean up the untracked files using 'git clean
<pathspec>', and then use a command chain involving 'git add -u
<pathspec>' and 'git apply' to clear the changes to files that are in
the index and were stashed.
As the <pathspec> only includes untracked files that were already
removed by 'git clean', the 'git add' call will barf, and so will 'git
apply', as there are no changes that need to be applied.
Fix this by avoiding the 'git clean' if a pathspec is given, and use the
pipeline that's used for pathspec mode to get rid of the untracked files
as well.
Reported-by: Marc Strapetz <marc.strapetz@syntevo.com> Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
An earlier change bba067d2 ("stash: don't delete untracked files
that match pathspec", 2018-01-06) was made by taking a suggestion in
a list discussion [1] but did not copy the suggested snippet
correctly. And the bug was unnoticed during the review and slipped
through.
In order to echo a tab character, it's better to use printf instead of
"echo -e", because it's more portable (for instance, "echo -e" doesn't work
as expected on a Mac).
This solves the "fatal: Not a valid object name" error in git-filter-branch
when using the --state-branch option.
Furthermore, let's switch from "/bin/echo" to just "echo", so that the
built-in echo command is used where available.
Signed-off-by: Michele Locati <michele@locati.it> Signed-off-by: Junio C Hamano <gitster@pobox.com>
On Windows, strftime() does not silently ignore invalid formats, but
warns about them and then returns 0 and sets errno to EINVAL.
Unfortunately, Git does not expect such a behavior, as it disagrees
with strftime()'s semantics on Linux. As a consequence, Git
misinterprets the return value 0 as "I need more space" and grows the
buffer. As the larger buffer does not fix the format, the buffer grows
and grows and grows until we are out of memory and abort.
Ideally, we would switch off the parameter validation just for
strftime(), but we cannot even override the invalid parameter handler
via _set_thread_local_invalid_parameter_handler() using MINGW because
that function is not declared. Even _set_invalid_parameter_handler(),
which *is* declared, does not help, as it simply does... nothing.
So let's just bite the bullet and override strftime() for MINGW and
abort on an invalid format string. While this does not provide the
best user experience, it is the best we can do.
See https://msdn.microsoft.com/en-us/library/fe06s4ak.aspx for more
details.
This fixes https://github.com/git-for-windows/git/issues/863
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
completion: complete tags with git tag --delete/--verify
Completion of tag names has worked for the short -d/-v options since 88e21dc746 ("Teach bash about completing arguments for git-tag",
2007-08-31). The long options were not added to "git tag" until many
years later, in c97eff5a95 ("git-tag: introduce long forms for the
options", 2011-08-28).
Extend tag name completion to --delete/--verify.
Signed-off-by: Todd Zullinger <tmz@pobox.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git shortlog cruft" aborted with a BUG message when run outside a
Git repository. The command has been taught to complain about
extra and unwanted arguments on its command line instead in such a
case.
The build procedure learned to optionally use symbolic links
(instead of hardlinks and copies) to install "git-foo" for built-in
commands, whose binaries are all identical.
* ab/install-symlinks:
Makefile: optionally symlink libexec/git-core binaries to bin/git
Makefile: add a gitexecdir_relative variable
Makefile: fix broken bindir_relative variable
Merge branch 'ml/filter-branch-no-op-error' into next
"git filter-branch" learned to use a different exit code to allow
the callers to tell the case where there was no new commits to
rewrite from other error cases.
* ml/filter-branch-no-op-error:
filter-branch: return 2 when nothing to rewrite
Merge branch 'jt/transfer-fsck-with-promissor' into next
The transfer.fsckobjects configuration tells "git fetch" to
validate the data and connected-ness of objects in the received
pack; the code to perform this check has been taught about the
narrow clone's convention that missing objects that are reachable
from objects in a pack that came from a promissor remote is OK.
* jt/transfer-fsck-with-promissor:
fetch-pack: do not check links for partial fetch
index-pack: support checking objects but not links
Git can be built to use either v1 or v2 of the PCRE library, and so
far, the build-time configuration USE_LIBPCRE=YesPlease instructed
the build procedure to use v1, but now it means v2. USE_LIBPCRE1
and USE_LIBPCRE2 can be used to explicitly choose which version to
use, as before.
* ab/pcre-v2:
Makefile: make USE_LIBPCRE=YesPlease mean v2, not v1
configure: detect redundant --with-libpcre & --with-libpcre1
configure: fix a regression in PCRE v1 detection
Merge branch 'ti/fetch-everything-local-optim' into next
A "git fetch" from a repository with insane number of refs into a
repository that is already up-to-date still wasted too many cycles
making many lstat(2) calls to see if these objects at the tips
exist as loose objects locally. These lstat(2) calls are optimized
away by enumerating all loose objects beforehand.
It is unknown if the new strategy negatively affects existing use
cases, fetching into a repository with many loose objects from a
repository with small number of refs.
* ti/fetch-everything-local-optim:
fetch-pack.c: use oidset to check existence of loose object
Merge branch 'nd/remove-ignore-env-field' into next
Code clean-up for the "repository" abstraction.
* nd/remove-ignore-env-field:
repository: delete ignore_env member
sha1_file.c: move delayed getenv(altdb) back to setup_git_env()
repository.c: delete dead functions
repository.c: move env-related setup code back to environment.c
repository: initialize the_repository in main()
"git rebase" learned "--recreate-merges" to transplant the whole
topology of commit graph elsewhere.
* js/rebase-recreate-merge:
rebase -i: introduce --recreate-merges=[no-]rebase-cousins
pull: accept --rebase=recreate to recreate the branch topology
sequencer: handle post-rewrite for merge commands
sequencer: make refs generated by the `label` command worktree-local
rebase: introduce the --recreate-merges option
rebase-helper --make-script: introduce a flag to recreate merges
sequencer: fast-forward merge commits, if possible
sequencer: introduce the `merge` command
sequencer: introduce new commands to reset the revision
git-rebase--interactive: clarify arguments
sequencer: make rearrange_squash() a bit more obvious
sequencer: avoid using errno clobbered by rollback_lock_file()
Merge branch 'ma/config-page-only-in-list-mode' into next
In a way similar to how "git tag" learned to honor the pager
setting only in the list mode, "git config" learned to ignore the
pager setting when it is used for setting values (i.e. when the
purpose of the operation is not to "show").
* ma/config-page-only-in-list-mode:
config: change default of `pager.config` to "on"
config: respect `pager.config` in list/get-mode only
t7006: add tests for how git config paginates
* sg/cvs-tests-with-x:
t9402-git-cvsserver-refs: don't check the stderr of a subshell
t9400-git-cvsserver-server: don't rely on the output of 'test_cmp'
* ab/perl-fixes:
perl Git::LoadCPAN: emit better errors under NO_PERL_CPAN_FALLBACKS
Makefile: add NO_PERL_CPAN_FALLBACKS knob
perl: move the perl/Git/FromCPAN tree to perl/FromCPAN
perl: generalize the Git::LoadCPAN facility
perl: move CPAN loader wrappers to another namespace
perl: update our copy of Mail::Address
perl: update our ancient copy of Error.pm
git-send-email: unconditionally use Net::{SMTP,Domain}
Git.pm: hard-depend on the File::{Temp,Spec} modules
gitweb: hard-depend on the Digest::MD5 5.8 module
Git.pm: add the "use warnings" pragma
Git.pm: remove redundant "use strict" from sub-package
perl: *.pm files should not have the executable bit
Makefile: optionally symlink libexec/git-core binaries to bin/git
Add a INSTALL_SYMLINKS option which if enabled, changes the default
hardlink installation method to one where the relevant binaries in
libexec/git-core are symlinked back to ../../bin, instead of being
hardlinked.
This new option also overrides the behavior of the existing
NO_*_HARDLINKS variables which in some cases would produce symlinks
within to libexec/, e.g. "git-add" symlinked to "git" which would be
copy of the "git" found in bin/, now "git-add" in libexec/ is always
going to be symlinked to the "git" found in the bin/ directory.
This option is being added because:
1) I think it makes what we're doing a lot more obvious. E.g. I'd
never noticed that the libexec binaries were really just hardlinks
since e.g. ls(1) won't show that in any obvious way. You need to
start stat(1)-ing things and look at the inodes to see what's
going on.
2) Some tools have very crappy support for hardlinks, e.g. the Git
shipped with GitLab is much bigger than it should be because
they're using a chef module that doesn't know about hardlinks, see
https://github.com/chef/omnibus/issues/827
I've also ran into other related issues that I think are explained
by this, e.g. compiling git with debugging and rpm refusing to
install a ~200MB git package with 2GB left on the FS, I think that
was because it doesn't consider hardlinks, just the sum of the
byte size of everything in the package.
As for the implementation, the "../../bin" noted above will vary given
some given some values of "../.." and "bin" depending on the depth of
the gitexecdir relative to the destdir, and the "bindir" target,
e.g. setting "bindir=/tmp/git/binaries gitexecdir=foo/bar/baz" will do
the right thing and produce this result:
$ file /tmp/git/foo/bar/baz/git-add
/tmp/git/foo/bar/baz/git-add: symbolic link to ../../../binaries/git
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This variable will be e.g. "libexec/git-core" if
gitexecdir=/tmp/git/libexec/git-core is given. It'll be used by a
subsequent change.
This is stolen from the yet-to-be integrated (needs resubmission)
"Makefile: add Perl runtime prefix support" patch on the mailing
list. See
<20180108030239.92036-3-dnj@google.com> (https://public-inbox.org/git/20180108030239.92036-3-dnj@google.com/).
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change the bindir_relative variable to work like the other *_relative
variables, which are computed as a function of the absolute
path. Before this change, supplying e.g. bindir=/tmp/git/binaries to
the Makefile would yield a bindir_relative of just "bin", as opposed
to "binaries".
This logic was originally added back in 026fa0d5ad ("Move computation
of absolute paths from Makefile to runtime (in preparation for
RUNTIME_PREFIX)", 2009-01-18), then later in 971f85388f ("Makefile:
make mandir, htmldir and infodir absolute", 2013-02-24) when
more *_relative variables were added those new variables didn't have
this bug, but bindir_relative was never fixed.
There is a small change in behavior here, which is that setting
bindir_relative as an argument to the Makefile won't work anymore, I
think that's fine, since this was always intended as an internal
variable (e.g. INSTALL documents bindir=*, not bindir_relative=*).
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
worktree prune: improve prune logic when worktree is moved
Automatic detection of worktree relocation by a user (via 'mv', for
instance) was removed by 618244e160 (worktree: stop supporting moving
worktrees manually - 2016-01-22). Prior to that,
.git/worktrees/<tag>/gitdir was updated whenever the worktree was
accessed in order to let the pruning logic know that the worktree was
"active" even if it disappeared for a while (due to being located on
removable media, for instance).
"git worktree move" has come so we don't really need this, but since
it's easy to do, perhaps we could keep supporting manual worktree move
a bit longer. Notice that when a worktree is active, the "index" file
should be updated pretty often in common case. The logic is updated to
check for index mtime to see if the worktree is alive.
The old logic of checking gitdir's mtime is dropped because nobody
updates it anyway. The new corner case is, if the index file does not
exist, we immediately remove the stale worktree. But if the "index"
file does not exist, you may have a bigger problem.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This "link" was a feature in early iterations of multiple worktree
functionality for some reason it was dropped [1]. Since nobody creates
this "link", there's no need to check it.
This is mostly used to let the user moves a worktree manually [2]. If
you move a worktree within the same file system, this hard link count
lets us know the worktree is still there even if we don't know where it
is.
We support 'worktree move' now and don't need this anymore.
[1] last appearance in v4 message-id: 1393675983-3232-25-git-send-email-pclouds@gmail.com
and the reason in v5 was "revisit later", message-id: 1394246900-31535-1-git-send-email-pclouds@gmail.com
[2] 23af91d102 (prune: strategies for linked checkouts - 2014-11-30)
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In order to be able to ship protocol v2 with only supporting fetch, we
need clients to not issue a request to use protocol v2 when pushing
(since the client currently doesn't know how to push using protocol v2).
This allows a client to have protocol v2 configured in
`protocol.version` and take advantage of using v2 for fetch and falling
back to using v0 when pushing while v2 for push is being designed.
We could run into issues if we didn't fall back to protocol v2 when
pushing right now. This is because currently a server will ignore a request to
use v2 when contacting the 'receive-pack' endpoint and fall back to
using v0, but when push v2 is rolled out to servers, the 'receive-pack'
endpoint will start responding using v2. So we don't want to get into a
state where a client is requesting to push with v2 before they actually
know how to push using v2.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach remote-curl the 'stateless-connect' command which is used to
establish a stateless connection with servers which support protocol
version 2. This allows remote-curl to act as a proxy, allowing the git
client to communicate natively with a remote end, simply using
remote-curl as a pass through to convert requests to http.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of always sending the Git-Protocol header with the configured
version with every http request, explicitly send it when discovering
refs and then only send it on subsequent http requests if the server
understood the version requested.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
remote-curl: store the protocol version the server responded with
Store the protocol version the server responded with when performing
discovery. This will be used in a future patch to either change the
'Git-Protocol' header sent in subsequent requests or to determine if a
client needs to fallback to using a different protocol version.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add the 'packet_buf_write_len()' function which allows for writing an
arbitrary length buffer into a 'struct strbuf' and formatting it in
packet-line format.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce the transport-helper capability 'stateless-connect'. This
capability indicates that the transport-helper can be requested to run
the 'stateless-connect' command which should attempt to make a
stateless connection with a remote end. Once established, the
connection can be used by the git client to communicate with
the remote end natively in a stateless-rpc manner as supported by
protocol v2. This means that the client must send everything the server
needs in a single request as the client must not assume any
state-storing on the part of the server or transport.
If a stateless connection cannot be established then the remote-helper
will respond in the same manner as the 'connect' command indicating that
the client should fallback to using the dumb remote-helper commands.
A future patch will implement the 'stateless-connect' capability in our
http remote-helper (remote-curl) so that protocol v2 can be used using
the http transport.
Signed-off-by: Brandon Williams <bmwill@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>