Amend code added in a6a8431968 (receive-pack.c: shorten the
execute_commands loop over all commands, 2015-01-07) and amended to
hard die in b6a4788586 (receive-pack.c: die instead of error in case
of possible future bug, 2015-01-07) to use the new bug() function
instead.
Let's also rename the warn_if_*() function that code is in to
BUG_if_*(), its name became outdated in b6a4788586.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change the assertions added in bf3ff338a2 (parse-options: stop
abusing 'callback' for lowlevel callbacks, 2019-01-27) to use optbug()
instead of BUG(). At this point we're looping over individual options,
so if we encounter any issues we'd like to report the offending option.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we run into bugs in parse-options.c usage it's good to be able to
note all the issues we ran into before dying. This use-case is why we
have the optbug() function introduced in 1e5ce570ca (parse-options:
clearer reporting of API misuse, 2010-12-02)
Let's change this code to use the new bug() API introduced in the
preceding commit, which cuts down on the verbosity of
parse_options_check().
There are existing uses of BUG() in adjacent code that should have
been using optbug() that aren't being changed here. That'll be done in
a subsequent commit. This only changes the optbug() callers.
Since this will invoke BUG() the previous exit(128) code will be
changed, but in this case that's what we want, i.e. to have
encountering a BUG() return the specific "BUG" exit code.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a bug() function to use in cases where we'd like to indicate a
runtime BUG(), but would like to defer the BUG() call because we're
possibly accumulating more bug() callers to exhaustively indicate what
went wrong.
We already have this sort of facility in various parts of the
codebase, just in the form of ad-hoc re-inventions of the
functionality that this new API provides. E.g. this will be used to
replace optbug() in parse-options.c, and the 'error("BUG:[...]' we do
in a loop in builtin/receive-pack.c.
Unlike the code this replaces we'll log to trace2 with this new bug()
function (as with other usage.c functions, including BUG()), we'll
also be able to avoid calls to xstrfmt() in some cases, as the bug()
function itself accepts variadic sprintf()-like arguments.
Any caller to bug() can follow up such calls with BUG_if_bug(),
which will BUG() out (i.e. abort()) if there were any preceding calls
to bug(), callers can also decide not to call BUG_if_bug() and leave
the resulting BUG() invocation until exit() time. There are currently
no bug() API users that don't call BUG_if_bug() themselves after a
for-loop, but allowing for not calling BUG_if_bug() keeps the API
flexible. As the tests and documentation here show we'll catch missing
BUG_if_bug() invocations in our exit() wrapper.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change the exit() wrapper added in ee4512ed48 (trace2: create new
combined trace facility, 2019-02-22) so that we'll split up the trace2
logging concerns from wanting to wrap the "exit()" function itself for
other purposes.
This makes more sense structurally, as we won't seem to conflate
non-trace2 behavior with the trace2 code. I'd previously added an
explanation for this in 368b584315 (common-main.c: call exit(), don't
return, 2021-12-07), that comment is being adjusted here.
Now the only thing we'll do if we're not using trace2 is to truncate
the "code" argument to the lowest 8 bits.
We only need to do that truncation on non-POSIX systems, but in
ee4512ed48 that "if defined(__MINGW32__)" code added in
47e3de0e79 (MinGW: truncate exit()'s argument to lowest 8 bits,
2009-07-05) was made to run everywhere. It might be good for clarify
to narrow that down by an "ifdef" again, but I'm not certain that in
the interim we haven't had some other non-POSIX systems rely the
behavior. On a POSIX system taking the lowest 8 bits is implicit, see
exit(3)[1] and wait(2)[2]. Let's leave a comment about that instead.
1. https://man7.org/linux/man-pages/man3/exit.3.html
2. https://man7.org/linux/man-pages/man2/wait.2.html
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
According to the HTML Standard FAQ:
“What is the DOCTYPE for modern HTML documents?
In text/html documents:
<!DOCTYPE html>
In documents delivered with an XML media type: no DOCTYPE is required
and its use is generally unnecessary. However, you may use one if you
want (see the following question). Note that the above is well-formed
XML.”
Source: [1]
Gitweb uses an XHTML 1.0 DOCTYPE:
<!DOCTYPE html PUBLIC
"-//W3C//DTD XHTML 1.0 Strict//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
While that DOCTYPE is still valid [2], it has several disadvantages:
1. It’s misleading. If an XML parser uses the DTD at the given link,
then the entities and ⋅ won’t get declared. Instead, the
parser has to use a DTD from the HTML Standard that has nothing to do
with XHTML 1.0 [2].
2. It’s obsolete. XHTML 1.0 was last revised in 2002 and was superseded in
2018 [3].
3. It’s unreliable. Gitweb uses and ⋅ but lets an external file
define them. “[…U]using entity references for characters in XML documents
is unsafe if they are defined in an external file (except for <, >,
&, ", and ').” [4]
[1]: <https://github.com/whatwg/html/blob/main/FAQ.md#what-is-the-doctype-for-modern-html-documents>
[2]: <https://html.spec.whatwg.org/multipage/xhtml.html#parsing-xhtml-documents>
[3]: <https://www.w3.org/TR/xhtml1/#xhtml>
[4]: <https://html.spec.whatwg.org/multipage/xhtml.html#writing-xhtml-documents>
Signed-off-by: Jason Yundt <jason@jasonyundt.email>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Branch names can't be empty, so config keys with an empty branch name,
e.g. "branch..remote", are silently ignored.
Since these config keys will never be useful, make it a fatal error when
remote.c finds a key that starts with "branch." and has an empty
subsection.
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
4a2dcb1a08 (remote: die if branch is not found in repository,
2021-11-17) introduced a regression where multiple config entries with
an empty branch name, e.g.
[branch ""]
remote = foo
merge = bar
could cause Git to fail when it tries to look up branch tracking
information.
We parse the config key to get (branch name, branch name length), but
when the branch name subsection is empty, we get a bogus branch name,
e.g. "branch..remote" gives (".remote", 0). We continue to use the bogus
branch name as if it were valid, and prior to 4a2dcb1a08, this wasn't an
issue because length = 0 caused the branch name to effectively be ""
everywhere.
However, that commit handles length = 0 inconsistently when we create
the branch:
- When find_branch() is called to check if the branch exists in the
branch hash map, it interprets a length of 0 to mean that it should
call strlen on the char pointer.
- But the code path that inserts into the branch hash map interprets a
length of 0 to mean that the string is 0-length.
This results in the bug described above:
- "branch..remote" looks for ".remote" in the branch hash map. Since we
do not find it, we insert the "" entry into the hash map.
- "branch..merge" looks for ".merge" in the branch hash map. Since we
do not find it, we again try to insert the "" entry into the hash map.
However, the entries in the branch hash map are supposed to be
appended to, not overwritten.
- Since overwriting an entry is a BUG(), Git fails instead of silently
ignoring the empty branch name.
Fix the bug by removing the convenience strlen functionality, so that
0 means that the string is 0-length. We still insert a bogus branch name
into the hash map, but this will be fixed in a later commit.
Reported-by: "Ing. Martin Prantl Ph.D." <perry@ntis.zcu.cz>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This reverts commit 9c539d1027 (config.mak.dev: alternative
workaround to gcc 12 warning in http.c, 2022-04-15).
Let's give GCC12's "dangling-pointer" warning a second chance, as we
have a more focused workaround for this particular compiler glitch.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fixes real problems noticed by gcc 12 and works around false
positives.
* js/ci-gcc-12-fixes:
dir.c: avoid "exceeds maximum object size" error with GCC v12.x
nedmalloc: avoid new compile error
compat/win32/syslog: fix use-after-realloc
As 'revert' and 'cherry-pick' share a lot of code, it is easy to
modify the behaviour of one command and inadvertently affect the
other. An earlier change to teach the '--reference' option and the
'revert.reference' configuration variable to the former was not
careful enough and 'cherry-pick --reference' wasn't rejected as an
error.
It is possible to think 'cherry-pick -x' might benefit from the
'--reference' option, but it is fundamentally different from
'revert' in at least two ways to make it questionable:
- 'revert' names a commit that is ancestor of the resulting commit,
so an abbreviated object name with human readable title is
sufficient to identify the named commit uniquely without using
the full object name. On the other hand, 'cherry-pick'
usually [*] picks a commit that is not an ancestor. It might be
even picking a private commit that never becomes part of the
public history.
- The whole commit message of 'cherry-pick' is a copy of the
original commit, and there is nothing gained to repeat only the
title part on 'cherry-picked from' message.
[*] well, you could revert and then you can pick the original that
was reverted to get back to where you were, but then you can
revert the revert to do the same thing.
Helped-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git add -i" was rewritten in C some time ago and has been in
testing; the reimplementation is now exposed to general public by
default.
* js/use-builtin-add-i:
add -i: default to the built-in implementation
t2016: require the PERL prereq only when necessary
The tests that ensured merges stop when interfering local changes
are present did not make sure that local changes are preserved; now
they do.
* jc/t6424-failing-merge-preserve-local-changes:
t6424: make sure a failed merge preserves local changes
With the new http.curloptResolve configuration, the CURLOPT_RESOLVE
mechanism that allows cURL based applications to use pre-resolved
IP addresses for the requests is exposed to the scripts.
* cc/http-curlopt-resolve:
http: add custom hostname to IP address resolutions
When operating at the scale that Scalar wants to support, certain data
shapes are more likely to cause undesirable performance issues, such as
large numbers of loose objects.
By including statistics about this, `scalar diagnose` now makes it
easier to identify such scenarios.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's helpful to see if there are other crud files in the pack
directory. Let's teach the `scalar diagnose` command to gather
file size information about pack files.
While at it, also enumerate the pack files in the alternate
object directories, if any are registered.
Signed-off-by: Matthew John Cheetham <mjcheetham@outlook.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When analyzing problems with large worktrees/repositories, it is useful
to know how close to a "full disk" situation Scalar/Git operates. Let's
include this information.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Over the course of Scalar's development, it became obvious that there is
a need for a command that can gather all kinds of useful information
that can help identify the most typical problems with large
worktrees/repositories.
The `diagnose` command is the culmination of this hard-won knowledge: it
gathers the installed hooks, the config, a couple statistics describing
the data shape, among other pieces of information, and then wraps
everything up in a tidy, neat `.zip` archive.
Note: originally, Scalar was implemented in C# using the .NET API, where
we had the luxury of a comprehensive standard library that includes
basic functionality such as writing a `.zip` file. In the C version, we
lack such a commodity. Rather than introducing a dependency on, say,
libzip, we slightly abuse Git's `archive` machinery: we write out a
`.zip` of the empty try, augmented by a couple files that are added via
the `--add-file*` options. We are careful trying not to modify the
current repository in any way lest the very circumstances that required
`scalar diagnose` to be run are changed by the `diagnose` run itself.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `scalar` command needs a Scalar enlistment for many subcommands, and
looks in the current directory for such an enlistment (traversing the
parent directories until it finds one).
These is subcommands can also be called with an optional argument
specifying the enlistment. Here, too, we traverse parent directories as
needed, until we find an enlistment.
However, if the specified directory does not even exist, or is not a
directory, we should stop right there, with an error message.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
By allowing the path to be enclosed in double-quotes, we can avoid
the limitation that paths cannot contain colons.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With the `--add-virtual-file=<path>:<content>` option, `git archive` now
supports use cases where relatively trivial files need to be added that
do not exist on disk.
This will allow us to generate `.zip` files with generated content,
without having to add said content to the object database and without
having to write it out to disk.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
[jc: tweaked <path> handling]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Pathspecs with only negative elements did not work with some
commands that pass the pathspec along to a subprocess. For
instance,
$ git add -p -- ':!*.txt'
should add everything except for paths ending in ".txt", but it gets
complaint from underlying "diff-index" and aborts.
We used to error out when a pathspec with only negative elements in
it, like the one in the above example. Later, 859b7f1d (pathspec:
don't error out on all-exclusionary pathspec patterns, 2017-02-07)
updated the logic to add an empty string as an extra element. The
intention was to let the extra element to match everything and let
the negative ones given by the user to subtract from it.
At around the same time, we were migrating from "an empty string is
a valid pathspec element that matches everything" to "either a dot
or ":/" is used to match all, and an empty string is rejected",
between d426430e (pathspec: warn on empty strings as pathspec,
2016-06-22) and 9e4e8a64 (pathspec: die on empty strings as
pathspec, 2017-06-06). I think 9e4e8a64, which happened long after
859b7f1d happened, was not careful enough to turn the empty string
859b7f1d added to either a dot or ":/".
A care should be taken as the definition of "everything" depends on
subcommand. For the purpose of "add -p", adding a "." to add
everything in the current directory is the right thing to do. But
for some other commands, ":/" (i.e. really really everything, even
things outside the current subdirectory) is the right choice.
We would break commands in a big way if we get this wrong, so add a
handful of test pieces to make sure the resulting code still
excludes the paths that are expected and includes "everything" else.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Document the interaction between --add-file and --prefix by giving an
example.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In http.c, the run_active_slot() function allows the given "slot" to
make progress by calling step_active_slots() in a loop repeatedly,
and the loop is not left until the request held in the slot
completes.
Ages ago, we used to use the slot->in_use member to get out of the
loop, which misbehaved when the request in "slot" completes (at
which time, the result of the request is copied away from the slot,
and the in_use member is cleared, making the slot ready to be
reused), and the "slot" gets reused to service a different request
(at which time, the "slot" becomes in_use again, even though it is
for a different request). The loop terminating condition mistakenly
thought that the original request has yet to be completed.
Today's code, after baa7b67d (HTTP slot reuse fixes, 2006-03-10)
fixed this issue, uses a separate "slot->finished" member that is
set in run_active_slot() to point to an on-stack variable, and the
code that completes the request in finish_active_slot() clears the
on-stack variable via the pointer to signal that the particular
request held by the slot has completed. It also clears the in_use
member (as before that fix), so that the slot itself can safely be
reused for an unrelated request.
One thing that is not quite clean in this arrangement is that,
unless the slot gets reused, at which point the finished member is
reset to NULL, the member keeps the value of &finished, which
becomes a dangling pointer into the stack when run_active_slot()
returns. Clear the finished member before the control leaves the
function, which has a side effect of unconfusing compilers like
recent GCC 12 that is over-eager to warn against such an assignment.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix memory leaks in transport_push(), where remote_refs and local_refs
are never freed.
116 bytes in 1 blocks are definitely lost in loss record 56 of 103
at 0x484486F: malloc (vg_replace_malloc.c:381)
by 0x4938D7E: strdup (strdup.c:42)
by 0x628418: xstrdup (wrapper.c:39)
by 0x4FD454: process_capabilities (connect.c:232)
by 0x4FD454: get_remote_heads (connect.c:354)
by 0x610A38: handshake (transport.c:333)
by 0x612B02: transport_push (transport.c:1302)
by 0x4803D6: push_with_options (push.c:357)
by 0x4811D6: do_push (push.c:414)
by 0x4811D6: cmd_push (push.c:650)
by 0x405210: run_builtin (git.c:465)
by 0x405210: handle_builtin (git.c:719)
by 0x406363: run_argv (git.c:786)
by 0x406363: cmd_main (git.c:917)
by 0x404F17: main (common-main.c:56)
5,912 (388 direct, 5,524 indirect) bytes in 2 blocks are definitely lost in loss record 98 of 103
at 0x4849464: calloc (vg_replace_malloc.c:1328)
by 0x628705: xcalloc (wrapper.c:150)
by 0x5C216D: alloc_ref_with_prefix (remote.c:975)
by 0x5C232A: alloc_ref (remote.c:983)
by 0x5C232A: one_local_ref (remote.c:2299)
by 0x5C232A: one_local_ref (remote.c:2289)
by 0x5BDB03: do_for_each_repo_ref_iterator (iterator.c:418)
by 0x5B4C4F: do_for_each_ref (refs.c:1486)
by 0x5B4C4F: refs_for_each_ref (refs.c:1492)
by 0x5B4C4F: for_each_ref (refs.c:1497)
by 0x5C6ADF: get_local_heads (remote.c:2310)
by 0x612A85: transport_push (transport.c:1286)
by 0x4803D6: push_with_options (push.c:357)
by 0x4811D6: do_push (push.c:414)
by 0x4811D6: cmd_push (push.c:650)
by 0x405210: run_builtin (git.c:465)
by 0x405210: handle_builtin (git.c:719)
by 0x406363: run_argv (git.c:786)
by 0x406363: cmd_main (git.c:917)
Signed-off-by: Frantisek Hrbata <frantisek@hrbata.com>
Reviewed-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It seems there is no reason to return 1 instead of -1 when push_refs()
is not set in transport vtable. Let's unify the error return values and
use the done label as a single exit point from transport_push().
Suggested-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Frantisek Hrbata <frantisek@hrbata.com>
Reviewed-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Remove the big indented block for transport_push() check in transport vtable
and let's just return error immediately. Hopefully this makes the code
more readable.
Signed-off-by: Frantisek Hrbata <frantisek@hrbata.com>
Reviewed-by: Josh Steadmon <steadmon@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A typical "git revert" commit uses the full title of the original
commit in its title, and starts its body of the message with:
This reverts commit 8fa7f667cf61386257c00d6e954855cc3215ae91.
This does not encourage the best practice of describing not just
"what" (i.e. "Revert X" on the title says what we did) but "why"
(i.e. and it does not say why X was undesirable).
We can instead phrase this first line of the body to be more like
This reverts commit 8fa7f667 (do this and that, 2022-04-25)
so that the title does not have to be
Revert "do this and that"
We can instead use the title to describe "why" we are reverting the
original commit.
Introduce the "--reference" option to "git revert", and also the
revert.reference configuration variable, which defaults to false, to
tweak the title and the first line of the draft commit message for
when creating a "revert" commit.
When this option is in use, the first line of the pre-filled editor
buffer becomes a comment line that tells the user to say _why_. If
the user exits the editor without touching this line by mistake,
what we prepare to become the first line of the body, i.e. "This
reverts commit 8fa7f667 (do this and that, 2022-04-25)", ends up to
be the title of the resulting commit. This behaviour is designed to
help such a user to identify such a revert in "git log --oneline"
easily so that it can be further reworded with "git rebase -i" later.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Refactor the tests that exercise implicit shutdown cases
to make them more robust and less racy.
The fsmonitor--daemon will implicitly shutdown in a variety
of situations, such as when the ".git" directory is deleted
or renamed.
The existing tests would delete or rename the directory, sleep
for one second, and then check the status of the daemon. This
is racy, since the client/status command has no way to sync
with the daemon. This was noticed occasionally on very slow
CI build machines where it would cause a random test to fail.
Replace the simple sleep with a sleep-and-retry loop.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create a test in t7527 to verify that we get a stray warning from
`git fsmonitor--daemon start` when indirectly called from
`git submodule absorbgitdirs`.
Update `git fsmonitor--daemon` to take (and ignore) the `--super-prefix`
argument to suppress the warning.
When we have:
1. a submodule with a `sub/.git/` directory (rather than a `sub/.git`
file).
2. `core.fsmonitor` is turned on in the submodule, but the daemon is
not yet started in the submodule.
3. and someone does a `git submodule absorbgitdirs` in the super.
Git will recursively invoke `git submodule--helper absorb-git-dirs`
in the submodule. This will read the index and may attempt to start
the fsmonitor--daemon with the `--super-prefix` argument.
`git fsmonitor--daemon start` does not accept the `--super-prefix`
argument and causes a warning to be issued.
This does not cause a problem because the `refresh_index()` code
assumes a trivial response if the daemon does not start.
The net-net is a harmelss, but stray warning. Lets eliminate the
warning.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Confirm that the daemon reports events using the on-disk
spelling for Unicode NFC/NFD characters. On APFS we still
have Unicode aliasing, so we cannot create two files that
only differ by NFC/NFD, but the on-disk format preserves
the spelling used to create the file. On HFS+ we also
have aliasing, but the path is always stored on disk in
NFD.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create a set of prereqs to help understand how file names
are handled by the filesystem when they contain NFC and NFD
Unicode characters.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Co-authored-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Emit NFC or NFC and NFD spellings of pathnames on macOS.
MacOS is Unicode composition insensitive, so NFC and NFD spellings are
treated as aliases and collide. While the spelling of pathnames in
filesystem events depends upon the underlying filesystem, such as
APFS, HFS+ or FAT32, the OS enforces such collisions regardless of
filesystem.
Teach the daemon to always report the NFC spelling and to report
the NFD spelling when stored in that format on the disk.
This is slightly more general than "core.precomposeUnicode".
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Test that FS events from the OS are received using the preserved,
on-disk spelling of files/directories rather than spelling used
to make the change.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Never set CE_FSMONITOR_VALID on the cache-entry of submodule
directories.
During a client command like 'git status', we may need to recurse
into each submodule to compute a status summary for the submodule.
Since the purpose of the ce_flag is to let Git avoid scanning a
cache-entry, setting the flag causes the recursive call to be
avoided and we report incorrect (no status) for the submodule.
We created an OS watch on the root directory of our working
directory and we receive events for everything in the cone
under it. When submodules are present inside our working
directory, we receive events for both our repo (the super) and
any subs within it. Since our index doesn't have any information
for items within the submodules, we can't use those events.
We could try to truncate the paths of those events back to the
submodule boundary and mark the GITLINK as dirty, but that
feels expensive since we would have to prefix compare every FS
event that we receive against a list of submodule roots. And
it still wouldn't be sufficient to correctly report status on
the submodule, since we don't have any space in the cache-entry
to cache the submodule's status (the 'SCMU' bits in porcelain
V2 speak). That is, the CE_FSMONITOR_VALID bit just says that
we don't need to scan/inspect it because we already know the
answer -- it doesn't say that the item is clean -- and we
don't have space in the cache-entry to store those answers.
So we should always do the recursive scan.
Therefore, we should never set the flag on GITLINK cache-entries.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create unit tests to move a directory. Verify that `git status`
gives the same result with and without FSMonitor enabled.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach Git to perform binary search over the cache-entries for a directory
notification and then linearly scan forward to find the immediate children.
Previously, when the FSMonitor reported a modified directory Git would
perform a linear search on the entire cache-entry array for all
entries matching that directory prefix and invalidate them. Since the
cache-entry array is already sorted, we can use a binary search to
find the first matching entry and then only linearly walk forward and
invalidate entries until the prefix changes.
Also, the original code would invalidate anything having the same
directory prefix. Since a directory event should only be received for
items that are immediately within the directory (and not within
sub-directories of it), only invalidate those entries and not the
whole subtree.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach the listener thread to shutdown the daemon if the spelling of the
worktree root directory changes.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Force shutdown fsmonitor daemon if the worktree root directory
is moved, renamed, or deleted.
Use Windows low-level GetFileInformationByHandle() to get and
compare the Windows system unique ID for the directory with a
cached version when we started up. This lets us detect the
case where someone renames the directory that we are watching
and then creates a new directory with the original pathname.
This is important because we are listening to a named pipe for
requests and they are stored in the Named Pipe File System (NPFS)
which a kernel-resident pseudo filesystem not associated with
the actual NTFS directory.
For example, if the daemon was watching "~/foo/", it would have
a directory-watch handle on that directory and a named-pipe
handle for "//./pipe/...foo". Moving the directory to "~/bar/"
does not invalidate the directory handle. (So the daemon would
actually be watching "~/bar" but listening on "//./pipe/...foo".
If the user then does "git init ~/foo" and causes another daemon
to start, the first daemon will still have ownership of the pipe
and the second daemon instance will fail to start. "git status"
clients in "~/foo" will ask "//./pipe/...foo" about changes and
the first daemon instance will tell them about "~/bar".
This commit causes the first daemon to shutdown if the system unique
ID for "~/foo" changes (changes from what it was when the daemon
started). Shutdown occurs after a periodic poll. After the
first daemon exits and releases the lock on the named pipe,
subsequent Git commands may cause another daemon to be started
on "~/foo". Similarly, a subsequent Git command may cause another
daemon to be started on "~/bar".
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extend the Windows version of the "health" thread to periodically
inspect the system and shutdown if warranted.
This commit updates the thread's wait loop to use a timeout and
defines a (currently empty) table of functions to poll the system.
A later commit will add functions to the table to actually
inspect the system.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create another thread to watch over the daemon process and
automatically shut it down if necessary.
This commit creates the basic framework for a "health" thread
to monitor the daemon and/or the file system. Later commits
will add platform-specific code to do the actual work.
The "health" thread is intended to monitor conditions that
would be difficult to track inside the IPC thread pool and/or
the file system listener threads. For example, when there are
file system events outside of the watched worktree root or if
we want to have an idle-timeout auto-shutdown feature.
This commit creates the health thread itself, defines the thread-proc
and sets up the thread's event loop. It integrates this new thread
into the existing IPC and Listener thread models.
This commit defines the API to the platform-specific code where all of
the monitoring will actually happen.
The platform-specific code for MacOS is just stubs. Meaning that the
health thread will immediately exit on MacOS, but that is OK and
expected. Future work can define MacOS-specific monitoring.
The platform-specific code for Windows sets up enough of the
WaitForMultipleObjects() machinery to watch for system and/or custom
events. Currently, the set of wait handles only includes our custom
shutdown event (sent from our other theads). Later commits in this
series will extend the set of wait handles to monitor other
conditions.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rename platform-specific listener thread related variables
and data types as we prepare to add another backend thread
type.
[] `struct fsmonitor_daemon_backend_data` becomes `struct fsm_listen_data`
[] `state->backend_data` becomes `state->listen_data`
[] `state->error_code` becomes `state->listen_error_code`
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Refactor daemon thread startup to make it easier to start
a third thread class to monitor the health of the daemon.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach the fsmonitor--daemon to CD outside of the worktree
before starting up.
The common Git startup mechanism causes the CWD of the daemon process
to be in the root of the worktree. On Windows, this causes the daemon
process to hold a locked handle on the CWD and prevents other
processes from moving or deleting the worktree while the daemon is
running.
CD to HOME before entering main event loops.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
