In 8abfdf44c8 (tests: explicitly use `git.exe` on Windows,
2018-11-14), we made sure to use the `.exe` file extension when
using an absolute path to `git.exe`, to avoid getting confused with a
file or directory in the same place that lacks said file extension.
For the same reason, we need to handle test-tool.exe the same way.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The optional 'Large Edge List' chunk of the commit graph file stores
parent information for commits with more than two parents, and the
names of most of the macros, variables, struct fields, and functions
related to this chunk contain the term "large edges", e.g.
write_graph_chunk_large_edges(). However, it's not a really great
term, as the edges to the second and subsequent parents stored in this
chunk are not any larger than the edges to the first and second
parents stored in the "main" 'Commit Data' chunk. It's the number of
edges, IOW number of parents, that is larger compared to non-merge and
"regular" two-parent merge commits. And indeed, two functions in
'commit-graph.c' have a local variable called 'num_extra_edges' that
refer to the same thing, and this "extra edges" term is much better at
describing these edges.
So let's rename all these references to "large edges" in macro,
variable, function, etc. names to "extra edges". There is a
GRAPH_OCTOPUS_EDGES_NEEDED macro as well; for the sake of consistency
rename it to GRAPH_EXTRA_EDGES_NEEDED.
We can do so safely without causing any incompatibility issues,
because the term "large edges" doesn't come up in the file format
itself in any form (the chunk's magic is {'E', 'D', 'G', 'E'}, there
is no 'L' in there), but only in the specification text. The string
"large edges", however, does come up in the output of 'git
commit-graph read' and in tests looking at its input, but that command
is explicitly documented as debugging aid, so we can change its output
and the affected tests safely.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Slightly optimize the "commit-graph write" step by using
FOR_EACH_OBJECT_PACK_ORDER with for_each_object_in_pack(). See commit
[1] and [2] for the facility and a similar optimization for "cat-file".
On Linux it is around 5% slower to run:
echo 1 >/proc/sys/vm/drop_caches &&
cat .git/objects/pack/* >/dev/null &&
git cat-file --batch-all-objects --batch-check --unordered
Than the same thing with the "cat" omitted. This is as expected, since
we're iterating in pack order and the "cat" is extra work.
Before this change the opposite was true of "commit-graph write". We
were 6% faster if we first ran "cat" to efficiently populate the FS
cache for our sole big pack on linux.git, than if we had populated it
via for_each_object_in_pack(). Now we're 3% faster without the "cat"
instead.
My tests were done on an unloaded Linux 3.10 system with 10 runs for
each. Derrick Stolee did his own tests on Windows[3] showing a 2%
improvement with a high degree of accuracy.
1. 736eb88fdc ("for_each_packed_object: support iterating in
pack-order", 2018-08-10)
2. 0750bb5b51 ("cat-file: support "unordered" output for
--batch-all-objects", 2018-08-10)
3. https://public-inbox.org/git/f71fa868-25e8-a9c9-46a6-611b987f1a8f@gmail.com/
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git gc" and "git repack" did not close the open packfiles that
they found unneeded before removing them, which didn't work on a
platform incapable of removing an open file. This has been
corrected.
* js/gc-repack-close-before-remove:
gc/repack: release packs when needed
The "--format=<placeholder>" option of for-each-ref, branch and tag
learned to show a few more traits of objects that can be learned by
the object_info API.
* ot/ref-filter-object-info:
ref-filter: give uintmax_t to format with %PRIuMAX
ref-filter: add docs for new options
ref-filter: add tests for deltabase
ref-filter: add deltabase option
ref-filter: add tests for objectsize:disk
ref-filter: add check for negative file size
ref-filter: add objectsize:disk option
Flaky tests can now be repeatedly run under load with the
"--stress" option.
* sg/stress-test:
test-lib: add the '--stress' option to run a test repeatedly under load
test-lib-functions: introduce the 'test_set_port' helper function
test-lib: set $TRASH_DIRECTORY earlier
test-lib: consolidate naming of test-results paths
test-lib: parse command line options earlier
test-lib: parse options in a for loop to keep $@ intact
test-lib: extract Bash version check for '-x' tracing
test-lib: translate SIGTERM and SIGHUP to an exit
The loose object cache used to optimize existence look-up has been
updated.
* rs/loose-object-cache-perffix:
object-store: retire odb_load_loose_cache()
object-store: use one oid_array per subdirectory for loose cache
object-store: factor out odb_clear_loose_cache()
object-store: factor out odb_loose_cache()
An inherently racy test that caused intermittent failures has been
removed.
* tg/t5570-drop-racy-test:
Revert "t/lib-git-daemon: record daemon log"
t5570: drop racy test
Earlier we added "-Wformat-security" to developer builds, assuming
that "-Wall" (which includes "-Wformat" which in turn is required
to use "-Wformat-security") is always in effect. This is not true
when config.mak.autogen is in use, unfortunately. This has been
fixed by unconditionally adding "-Wall" to developer builds.
* jk/dev-build-format-security:
config.mak.dev: add -Wall, primarily for -Wformat, to help autoconf users
"git cherry-pick -m1" was forbidden when picking a non-merge
commit, even though there _is_ parent number 1 for such a commit.
This was done to avoid mistakes back when "cherry-pick" was about
picking a single commit, but is no longer useful with "cherry-pick"
that can pick a range of commits. Now the "-m$num" option is
allowed when picking any commit, as long as $num names an existing
parent of the commit.
Technically this is a backward incompatible change; hopefully
nobody is relying on the error-checking behaviour.
* so/cherry-pick-always-allow-m1:
t3506: validate '-m 1 -ff' is now accepted for non-merge commits
t3502: validate '-m 1' argument is now accepted for non-merge commits
cherry-pick: do not error on non-merge commits when '-m 1' is specified
t3510: stop using '-m 1' to force failure mid-sequence of cherry-picks
"git worktree remove" and "git worktree move" refused to work when
there is a submodule involved. This has been loosened to ignore
uninitialized submodules.
* nd/worktree-remove-with-uninitialized-submodules:
worktree: allow to (re)move worktrees with uninitialized submodules
The test suite tried to see if it is run under bash, but the check
itself failed under some other implementations of shell (notably
under NetBSD). This has been corrected.
* sg/test-bash-version-fix:
test-lib: check Bash version for '-x' without using shell arrays
Portability updates for the HPE NonStop platform.
* rb/hpe:
compat/regex/regcomp.c: define intptr_t and uintptr_t on NonStop
git-compat-util.h: add FLOSS headers for HPE NonStop
config.mak.uname: support for modern HPE NonStop config.
transport-helper: drop read/write errno checks
transport-helper: use xread instead of read
With zsh, "git cmd path<TAB>" was completed to "git cmd path name"
when the completed path has a special character like SP in it,
without any attempt to keep "path name" a single filename. This
has been fixed to complete it to "git cmd path\ name" just like
Bash completion does.
* cy/zsh-completion-SP-in-path:
completion: treat results of git ls-tree as file paths
zsh: complete unquoted paths with spaces correctly
The core.worktree setting in a submodule repository should not be
pointing at a directory when the submodule loses its working tree
(e.g. getting deinit'ed), but the code did not properly maintain
this invariant.
* sb/submodule-unset-core-worktree-when-worktree-is-lost:
submodule deinit: unset core.worktree
submodule--helper: fix BUG message in ensure_core_worktree
submodule: unset core.worktree if no working tree is present
submodule update: add regression test with old style setups
Some of the documentation pages formatted incorrectly with
Asciidoctor, which have been fixed.
* ma/asciidoctor:
git-status.txt: render tables correctly under Asciidoctor
Documentation: do not nest open blocks
git-column.txt: fix section header
"git stripspace" should be usable outside a git repository, but
under the "-s" or "-c" mode, it didn't.
* jn/stripspace-wo-repository:
stripspace: allow -s/-c outside git repository
"git submodule update" ought to use a single job unless asked, but
by mistake used multiple jobs, which has been fixed.
* sb/submodule-fetchjobs-default-to-one:
submodule update: run at most one fetch job unless otherwise set
The MSYS2 runtime does its best to emulate the command-line wildcard
expansion and de-quoting which would be performed by the calling Unix
shell on Unix systems.
Those Unix shell quoting rules differ from the quoting rules applying to
Windows' cmd and Powershell, making it a little awkward to quote
command-line parameters properly when spawning other processes.
In particular, git.exe passes arguments to subprocesses that are *not*
intended to be interpreted as wildcards, and if they contain
backslashes, those are not to be interpreted as escape characters, e.g.
when passing Windows paths.
Note: this is only a problem when calling MSYS2 executables, not when
calling MINGW executables such as git.exe. However, we do call MSYS2
executables frequently, most notably when setting the use_shell flag in
the child_process structure.
There is no elegant way to determine whether the .exe file to be
executed is an MSYS2 program or a MINGW one. But since the use case of
passing a command line through the shell is so prevalent, we need to
work around this issue at least when executing sh.exe.
Let's introduce an ugly, hard-coded test whether argv[0] is "sh", and
whether it refers to the MSYS2 Bash, to determine whether we need to
quote the arguments differently than usual.
That still does not fix the issue completely, but at least it is
something.
Incidentally, this also fixes the problem where `git clone \\server\repo`
failed due to incorrect handling of the backslashes when handing the path
to the git-upload-pack process.
Further, we need to take care to quote not only whitespace and
backslashes, but also curly brackets. As aliases frequently go through
the MSYS2 Bash, and as aliases frequently get parameters such as
HEAD@{yesterday}, this is really important. As an early version of this
patch broke this, let's make sure that this does not regress by adding a
test case for that.
Helped-by: Kim Gybels <kgybels@infogroep.be>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Due to a quirk in Git's method to spawn git-upload-pack, there is a
problem when passing paths with backslashes in them: Git will force the
command-line through the shell, which has different quoting semantics in
Git for Windows (being an MSYS2 program) than regular Win32 executables
such as git.exe itself.
The symptom is that the first of the two backslashes in UNC paths of the
form \\myserver\folder\repository.git is *stripped off*.
Document this bug by introducing a test case.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a user fetches refs/heads/master from a repo with namespace "ns", the
remote is expected to (1) not send the real refs/heads/master, and (2)
send refs/namespaces/ns/refs/heads/master with the name
refs/heads/master. (1) indeed happens now, but not (2) - Git only sends
refs that have the user-given prefix, but it checks them against the
full name of the ref (the one starting with refs/namespaces), and not
the namespace-stripped one.
This is demonstrated by the patch in the test. Currently, it results in
"fatal: couldn't find remote ref refs/heads/master" despite both
unnamespaced and namespaced master being present. With the code change,
it produces the expected result.
Check the ref prefixes against the namespace-stripped name.
This bug was discovered through applying patches [1] that override
protocol.version to 2 in repositories when running tests, allowing us to
notice differences in behavior across different protocol versions.
[1] https://public-inbox.org/git/cover.1547677183.git.jonathantanmy@google.com/
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the file system is case-insensitive, we really must be careful to
ignore differences in case only.
This fixes https://github.com/git-for-windows/git/issues/735
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Perforce requires a complete list of files being operated on. If
git is updating an existing shelved changelist, then any files
which are moved or copied were not being added to this list.
Signed-off-by: Luke Diamand <luke@diamand.org>
Acked-by: Andrey Mazo <amazo@checkvideo.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Updating a shelved P4 changelist where one or more files have
been moved or copied does not work. Add a test for this.
The problem is that P4 requires a complete list of the files being
changed, and move/copy only includes the _source_ in the case of
updating a shelved changelist. This results in errors from Perforce
such as:
//depot/src - needs tofile //depot/dst
Submit aborted -- fix problems then use 'p4 submit -c 1234'
Signed-off-by: Luke Diamand <luke@diamand.org>
Acked-by: Andrey Mazo <amazo@checkvideo.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 9472935d81 (add: introduce "--renormalize", 2017-11-16) taught
git-add to pass HASH_RENORMALIZE to add_to_index(), which then passes
the flag along to index_path(). However, the flags taken by
add_to_index() and the ones taken by index_path() are distinct
namespaces. We cannot take HASH_* flags in add_to_index(), because they
overlap with the ADD_CACHE_* flags we already take (in this case,
HASH_RENORMALIZE conflicts with ADD_CACHE_IGNORE_ERRORS).
We can solve this by adding a new ADD_CACHE_RENORMALIZE flag, and using
it to set HASH_RENORMALIZE within add_to_index(). In order to make it
clear that these two flags come from distinct sets, let's also change
the name "newflags" in the function to "hash_flags".
Reported-by: Dmitriy Smirnov <dmitriy.smirnov@jetbrains.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When Git determines whether a file has changed, it looks at the mtime,
at the file size, and to detect changes even if the mtime is the same
(on Windows, the mtime granularity is 100ns, read: if two files are
written within the same 100ns time slot, they have the same mtime) and
even if the file size is the same, Git also looks at the inode/device
numbers.
This design obviously comes from a Linux background, where `lstat()`
calls were designed to be cheap.
On Windows, there is no `lstat()`. It has to be emulated. And while
obtaining the mtime and the file size is not all that expensive (you can
get both with a single `GetFileAttributesW()` call), obtaining the
equivalent of the inode and device numbers is very expensive (it
requires a call to `GetFileInformationByHandle()`, which in turn
requires a file handle, which is *a lot* more expensive than one might
imagine).
As it is very uncommon for developers to modify files within 100ns time
slots, Git for Windows chooses not to fill inode/device numbers
properly, but simply sets them to 0.
However, in t6042 the files file_v1 and file_v2 are typically written
within the same 100ns time slot, and they do not differ in file size. So
the minor modification is not picked up.
Let's work around this issue by avoiding the `git mv` calls in the
'mod6-setup: chains of rename/rename(1to2) and rename/rename(2to1)' test
case. The target files are overwritten anyway, so it is not like we
really rename those files. This fixes the issue because `git add` will
now add the files as new files (as opposed to existing, just renamed
files).
Functionally, we do not change anything because we replace two `git mv
<old> <new>` calls (where `<new>` is completely overwritten and `git
add`ed later anyway) by `git rm <old>` calls (removing other files, too,
that are also completely overwritten and `git add`ed later).
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Define a GIT_TEST_SIDEBAND_ALL environment variable meant to be used
from tests. When set to true, this overrides uploadpack.allowsidebandall
to true, allowing the entire test suite to be run as if this
configuration is in place for all repositories.
As of this patch, all tests pass whether GIT_TEST_SIDEBAND_ALL is unset
or set to 1.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, a response to a fetch request has sideband support only while
the packfile is being sent, meaning that the server cannot send notices
until the start of the packfile.
Extend sideband support in protocol v2 fetch responses to the whole
response. upload-pack will advertise it if the
uploadpack.allowsidebandall configuration variable is set, and
fetch-pack will automatically request it if advertised.
If the sideband is to be used throughout the whole response, upload-pack
will use it to send errors instead of prefixing a PKT-LINE payload with
"ERR ".
This will be tested in a subsequent patch.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A subsequent patch will teach struct packet_reader a new field that, if
set, instructs it to interpret read data as multiplexed. This will
create a dependency from pkt-line to sideband.
To avoid a circular dependency, split recv_sideband() into 2 parts: the
reading loop (left in recv_sideband()) and the processing of the
contents (in demultiplex_sideband()), and move the former into pkt-line.
This reverses the direction of dependency: sideband no longer depends on
pkt-line, and pkt-line now depends on sideband.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
GCC 8 introduced the new -Wcast-function-type warning, which is
implied by -Wextra (which, in turn is enabled in our DEVELOPER flags).
When building Git with GCC 8 and this warning enabled on a non-glibc
platform [1], one is greeted with a screenful of compiler
warnings/errors:
compat/obstack.c: In function '_obstack_begin':
compat/obstack.c:162:17: error: cast between incompatible function types from 'void * (*)(long int)' to 'struct _obstack_chunk * (*)(void *, long int)' [-Werror=cast-function-type]
h->chunkfun = (struct _obstack_chunk * (*)(void *, long)) chunkfun;
^
compat/obstack.c:163:16: error: cast between incompatible function types from 'void (*)(void *)' to 'void (*)(void *, struct _obstack_chunk *)' [-Werror=cast-function-type]
h->freefun = (void (*) (void *, struct _obstack_chunk *)) freefun;
^
compat/obstack.c:116:8: error: cast between incompatible function types from 'struct _obstack_chunk * (*)(void *, long int)' to 'struct _obstack_chunk * (*)(long int)' [-Werror=cast-function-type]
: (*(struct _obstack_chunk *(*) (long)) (h)->chunkfun) ((size)))
^
compat/obstack.c:168:22: note: in expansion of macro 'CALL_CHUNKFUN'
chunk = h->chunk = CALL_CHUNKFUN (h, h -> chunk_size);
^~~~~~~~~~~~~
<snip>
'struct obstack' stores pointers to two functions to allocate and free
"chunks", and depending on how obstack is used, these functions take
either one parameter (like standard malloc() and free() do; this is
how we use it in 'kwset.c') or two parameters. Presumably to reduce
memory footprint, a single field is used to store the function pointer
for both signatures, and then it's casted to the appropriate signature
when the function pointer is accessed. These casts between function
pointers with different number of parameters are what trigger those
compiler errors.
Modify 'struct obstack' to use unions to store function pointers with
different signatures, and then use the union member with the
appropriate signature when accessing these function pointers. This
eliminates the need for those casts, and thus avoids this compiler
error.
[1] Compiling 'compat/obstack.c' on a platform with glibc is sort of
a noop, see the comment before '# define ELIDE_CODE', so this is
not an issue on common Linux distros.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It looks like it is a copy-paste error made in 80f2a6097c
(t/helper: add test-ref-store to test ref-store functions,
2017-03-26) to pass "old-sha1" instead of "new-sha1" to
notnull() when we get the new sha1 argument from
const char **argv.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Acked-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The comment explaining how to build the fuzzers was broken in
927c77e7d4 ("Makefile: use FUZZ_CXXFLAGS for linking fuzzers",
2018-11-14).
When building fuzzers, all .c files must be compiled with coverage
tracing enabled. This is not possible when using only FUZZ_CXXFLAGS, as
that flag is only applied to the fuzzers themselves. Switching back to
CFLAGS fixes the issue.
Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
fuzz-commit-graph identified a case where Git will read past the end of
a buffer containing a commit graph if the graph's header has an
incorrect chunk count. A simple bounds check in parse_commit_graph()
prevents this.
Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Break load_commit_graph_one() into a new function, parse_commit_graph().
The latter function operates on arbitrary buffers, which makes it
suitable as a fuzzing target. Since parse_commit_graph() is only called
by load_commit_graph_one() (and the fuzzer described below), we omit
error messages that would be duplicated by the caller.
Adds fuzz-commit-graph.c, which provides a fuzzing entry point
compatible with libFuzzer (and possibly other fuzzing engines).
Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When communicating with a remote server or a subprocess, use
expanded numbers rather than numbers with scaling suffix in the
object filter spec (e.g. "limit:blob=1k" becomes
"limit:blob=1024").
Update the protocol docs to note that clients should always perform this
expansion, to allow for more compatibility between server
implementations.
Signed-off-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a tree has already been recorded as omitted, we don't need to
traverse it again just to collect its omits. Stop traversing trees a
second time when collecting omits.
Signed-off-by: Matthew DeVore <matvore@google.com>
Reviewed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Implement positive values for <depth> in the tree:<depth> filter. The
exact semantics are described in Documentation/rev-list-options.txt.
The long-term goal at the end of this is to allow a partial clone to
eagerly fetch an entire directory of files by fetching a tree and
specifying <depth>=1. This, for instance, would make a build operation
fast and convenient. It is fast because the partial clone does not need
to fetch each file individually, and convenient because the user does
not need to supply a sparse-checkout specification.
Another way of considering this feature is as a way to reduce
round-trips, since the client can get any number of levels of
directories in a single request, rather than wait for each level of tree
objects to come back, whose entries are used to construct a new request.
Signed-off-by: Matthew DeVore <matvore@google.com>
Reviewed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
