Junio C Hamano [Wed, 19 May 2021 23:54:59 +0000 (08:54 +0900)]
Merge branch 'ew/sha256-clone-remote-curl-fix'
"git clone" from SHA256 repository by Git built with SHA-1 as the
default hash algorithm over the dumb HTTP protocol did not
correctly set up the resulting repository, which has been corrected.
* ew/sha256-clone-remote-curl-fix:
remote-curl: fix clone on sha256 repos
Junio C Hamano [Wed, 19 May 2021 23:54:58 +0000 (08:54 +0900)]
Merge branch 'en/dir-traversal'
"git clean" and "git ls-files -i" had confusion around working on
or showing ignored paths inside an ignored directory, which has
been corrected.
* en/dir-traversal:
dir: introduce readdir_skip_dot_and_dotdot() helper
dir: update stale description of treat_directory()
dir: traverse into untracked directories if they may have ignored subfiles
dir: avoid unnecessary traversal into ignored directory
t3001, t7300: add testcase showcasing missed directory traversal
t7300: add testcase showing unnecessary traversal into ignored directory
ls-files: error out on -i unless -o or -c are specified
dir: report number of visited directories and paths with trace2
dir: convert trace calls to trace2 equivalents
Han-Wen Nienhuys [Wed, 19 May 2021 15:31:28 +0000 (15:31 +0000)]
refs: make explicit that ref_iterator_peel returns boolean
Use -1 as error return value throughout.
This removes spurious differences in the GIT_TRACE_REFS output, depending on the
ref storage backend active.
Before, the cached ref_iterator (but only that iterator!) would return
peel_object() output directly. No callers relied on the peel_status values
beyond success/failure. All calls to these functions go through
peel_iterated_oid(), which returns peel_object() as a fallback, but also
squashing the error values.
The iteration interface already passes REF_ISSYMREF and REF_ISBROKEN through the
flags argument, so the additional error values in enum peel_status provide no
value.
The ref iteration interface provides a separate peel() function because certain
formats (eg. packed-refs and reftable) can store the peeled object next to the
tag SHA1. Passing the peeled SHA1 as an optional argument to each_ref_fn maps
more naturally to the implementation of ref databases. Changing the code in this
way is left for a future refactoring.
Signed-off-by: Han-Wen Nienhuys <hanwen@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 19 May 2021 16:11:05 +0000 (12:11 -0400)]
fetch-pack: signal v2 server that we are done making requests
When fetching with the v0 protocol over ssh (or a local upload-pack with
pipes), the server closes the connection as soon as it is finished
sending the pack. So even though the client may still be operating on
the data via index-pack (e.g., resolving deltas, checking connectivity,
etc), the server has released all resources.
With the v2 protocol, however, the server considers the ssh session only
as a transport, with individual requests coming over it. After sending
the pack, it goes back to its main loop, waiting for another request to
come from the client. As a result, the ssh session hangs around until
the client process ends, which may be much later (because resolving
deltas, etc, may consume a lot of CPU).
This is bad for two reasons:
- it's consuming resources on the server to leave open a connection
that won't see any more use
- if something bad happens to the ssh connection in the meantime (say,
it gets killed by the network because it's idle, as happened in a
real-world report), then ssh will exit non-zero, and we'll propagate
the error up the stack.
The server is correct here not to hang up after serving the pack. The v2
protocol's design is meant to allow multiple requests like this, and
hanging up would be the wrong thing for a hypothetical client which was
planning to make more requests (though in practice, the git.git client
never would, and I doubt any other implementations would either).
The right thing is instead for the client to signal to the server that
it's not interested in making more requests. We can do that by closing
the pipe descriptor we use to write to ssh. This will propagate to the
server upload-pack as an EOF when it tries to read the next request (and
then it will close its half, and the whole connection will go away).
It's important to do this "half duplex" shutdown, because we have to do
it _before_ we actually receive the pack. This is an artifact of the way
fetch-pack and index-pack (or unpack-objects) interact. We hand the
connection off to index-pack (really, a sideband demuxer which feeds
it), and then wait until it returns. And it doesn't do that until it has
resolved all of the deltas in the pack, even though it was done reading
from the server long before.
So just closing the connection fully after index-pack returns would be
too late; we'd have held it open much longer than was necessary. And
teaching index-pack to close the connection is awkward. It's not even
seeing the whole conversation (the sideband demuxer is, but it doesn't
actually know what's in the packets, or when the end comes).
Note that this close() is happening deep within the transport code. It's
possible that a caller would want to perform other operations over the
same ssh transport after receiving the pack. But as of the current code,
none of the callers do, and there haven't been discussions of any plans
to change this. If we need to support that later, we can probably do so
by passing down a flag for "you're the last request on the transport;
it's OK to close" instead of the code just assuming that's true.
The description above all discusses v2 ssh, so it's worth thinking about
how this interacts with other protocols:
- in v0 protocols, we could do the same half-duplex shutdown (it just
goes into the v0 do_fetch_pack() instead). This does work, but since
it doesn't have the same persistence problem in the first place,
there's little reason to change it at this point.
- local fetches against git-upload-pack on the same machine will
behave the same as ssh (they are talking over two pipes, and see EOF
on their input pipe)
- fetches against git-daemon will run this same code, and close one of
the descriptors. In practice, this won't do anything, since there
our two descriptors are dups of each other, and not part of a
half-duplex pair. The right thing would probably be to call
shutdown(SHUT_WR) on it. I didn't bother with that here. It doesn't
face the same error-code problem (since it's just a TCP connection),
so it's really only an optimization problem. And git:// is not that
widely used these days, and has less impact on server resources than
an ssh termination.
- v2 http doesn't suffer from this problem in the first place, as our
pipes terminate at a local git-remote-https, which is passing data
along as individual requests via curl. Probably curl is keeping the
TCP/TLS connection open for more requests, and we might be able to
tell it manually "hey, we are done making requests now". But I think
that's much less important. It again doesn't suffer from the
error-code problem, and HTTP keepalive is pretty well understood
(importantly, the timeouts can be set low, because clients like curl
know how to reconnect for subsequent requests if necessary). So it's
probably not worth figuring out how to tell curl that we're done
(though if we do, this patch is probably the first step anyway;
fetch-pack closes the pipe back to remote-https, which would be the
signal that it should tell curl we're done).
The code is pretty straightforward. We close the pipe at the right
moment, and set it to -1 to mark it as invalid. I modified the later
cleanup code to avoid calling close(-1). That's not strictly necessary,
since close(-1) is a noop, but hopefully makes things a bit more obvious
to a reader.
I suspect that trying to call more transport functions after the close()
(e.g., calling transport_fetch_refs() again) would fail, as it's not
smart enough to realize we need to re-open the ssh connection. But
that's already true when v0 is in use. And no current callers want to do
that (and again, the solution is probably a flag in the transport code
to keep things open, which can be added later).
There's no test here, as the situation it covers is inherently racy (the
question is when upload-pack exits, compared to when index-pack finishes
resolving deltas and exits). The rather gross shell snippet below does
recreate the problematic situation; when run on a sufficiently-large
repository (git.git works fine), it kills an "idle" upload-pack while
the client is resolving deltas, leading to a failed clone.
(
git clone --no-local --progress . foo.git 2>&1
echo >&2 "clone exit code=$?"
) |
tr '\r' '\n' |
while read line
do
case "$done,$line" in
,Resolving*)
echo "hit resolving deltas; killing upload-pack"
killall -9 git-upload-pack
done=t
;;
esac
done
Reported-by: Greg Pflaum <greg.pflaum@pnp-hcl.com> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Wed, 19 May 2021 11:17:15 +0000 (07:17 -0400)]
clone: clean up directory after transport_fetch_refs() failure
git-clone started respecting errors from the transport subsystem in aab179d937 (builtin/clone.c: don't ignore transport_fetch_refs() errors,
2020-12-03). However, that commit didn't handle the cleanup of the
filesystem quite right.
The cleanup of the directory that cmd_clone() creates is done by an
atexit() handler, which we control with a flag. It starts as
JUNK_LEAVE_NONE ("clean up everything"), then progresses to
JUNK_LEAVE_REPO when we know we have a valid repo but not working tree,
and then finally JUNK_LEAVE_ALL when we have a successful checkout.
Most errors cause us to die(), which then triggers the handler to do the
right thing based on how far into cmd_clone() we got. But the checks
added by aab179d937 instead set the "err" variable and then jump to a
new "cleanup" label, which then returns our non-zero status. However,
the code after the cleanup label includes setting the flag to
JUNK_LEAVE_ALL, and so we accidentally leave the repository and working
tree in place.
One obvious option to fix this is to reorder the end of the function to
set the flag first, before cleanup code, and put the label between them.
But we can observe another small bug: the error return from
transport_fetch_refs() is generally "-1", and we propagate that to the
return value of cmd_clone(), which ultimately becomes the exit code of
the process. And we try to avoid transmitting negative values via exit
codes (only the low 8 bits are passed along as an unsigned value, though
in practice for "-1" this at least retains the property that it's
non-zero).
Instead, let's just die(). That makes us consistent with rest of the
code in the function. It does add a new "fatal:" line to the output, but
I'd argue that's a good thing:
- in the rare case that the transport code didn't say anything, now
the user gets _some_ error message
- even if the transport code said something like "error: ssh died of
signal 9", it's nice to also say "fatal" to indicate that we
considered that to be a show-stopper.
Triggering this in the test suite turns out to be surprisingly
difficult. Almost every error we'd encounter, including ones deep inside
the transport code, cause us to just die() right there! However, one way
is to put a fake wrapper around git-upload-pack that sends the complete
packfile but exits with a failure code.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee [Tue, 18 May 2021 18:32:48 +0000 (18:32 +0000)]
read-cache: delete unused hashing methods
These methods were marked as MAYBE_UNUSED in the previous change to
avoid a complicated diff. Delete them entirely, since we now use the
hashfile API instead of this custom hashing code.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee [Tue, 18 May 2021 18:32:47 +0000 (18:32 +0000)]
read-cache: use hashfile instead of git_hash_ctx
The do_write_index() method in read-cache.c has its own hashing logic
and buffering mechanism. Specifically, the ce_write() method was
introduced by 4990aadc (Speed up index file writing by chunking it
nicely, 2005-04-20) and similar mechanisms were introduced a few months
later in c38138cd (git-pack-objects: write the pack files with a SHA1
csum, 2005-06-26). Based on the timing, in the early days of the Git
codebase, I figured that these roughly equivalent code paths were never
unified only because it got lost in the shuffle. The hashfile API has
since been used extensively in other file formats, such as pack-indexes,
multi-pack-indexes, and commit-graphs. Therefore, it seems prudent to
unify the index writing code to use the same mechanism.
I discovered this disparity while trying to create a new index format
that uses the chunk-format API. That API uses a hashfile as its base, so
it is incompatible with the custom code in read-cache.c.
This rewrite is rather straightforward. It replaces all writes to the
temporary file with writes to the hashfile struct. This takes care of
many of the direct interactions with the_hash_algo.
There are still some git_hash_ctx uses remaining: the extension headers
are hashed for use in the End of Index Entries (EOIE) extension. This
use of the git_hash_ctx is left as-is. There are multiple reasons to not
use a hashfile here, including the fact that the data is not actually
writing to a file, just a hash computation. These hashes do not block
our adoption of the chunk-format API in a future change to the index, so
leave it as-is.
The internals of the algorithms are mostly identical. Previously, the
hashfile API used a smaller 8KB buffer instead of the 128KB buffer from
read-cache.c. The previous change already unified these sizes.
There is one subtle point: we do not pass the CSUM_FSYNC to the
finalize_hashfile() method, which differs from most consumers of the
hashfile API. The extra fsync() call indicated by this flag causes a
significant peformance degradation that is noticeable for quick
commands that write the index, such as "git add". Other consumers can
absorb this cost with their more complicated data structure
organization, and further writing structures such as pack-files and
commit-graphs is rarely in the critical path for common user
interactions.
Some static methods become orphaned in this diff, so I marked them as
MAYBE_UNUSED. The diff is much harder to read if they are deleted during
this change. Instead, they will be deleted in the following change.
In addition to the test suite passing, I computed indexes using the
previous binaries and the binaries compiled after this change, and found
the index data to be exactly equal. Finally, I did extensive performance
testing of "git update-index --force-write" on repos of various sizes,
including one with over 2 million paths at HEAD. These tests
demonstrated less than 1% difference in behavior. As expected, the
performance should be considered unchanged. The previous changes to
increase the hashfile buffer size from 8K to 128K ensured this change
would not create a peformance regression.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee [Tue, 18 May 2021 18:32:46 +0000 (18:32 +0000)]
csum-file.h: increase hashfile buffer size
The hashfile API uses a hard-coded buffer size of 8KB and has ever since
it was introduced in c38138c (git-pack-objects: write the pack files
with a SHA1 csum, 2005-06-26). It performs a similar function to the
hashing buffers in read-cache.c, but that code was updated from 8KB to
128KB in f279894 (read-cache: make the index write buffer size 128K,
2021-02-18). The justification there was that do_write_index() improves
from 1.02s to 0.72s. Since our end goal is to have the index writing
code use the hashfile API, we need to unify this buffer size to avoid a
performance regression.
There is a buffer, 'check_buffer', that is used to verify the check_fd
file descriptor. When this buffer increases to 128K to fit the data
being flushed, it causes the stack to overflow the limits placed in the
test suite. To avoid issues with stack size, move both 'buffer' and
'check_buffer' to be heap pointers within 'struct hashfile'. The
'check_buffer' member is left as NULL unless check_fd is set in
hashfd_check(). Both buffers are cleared as part of finalize_hashfile()
which also frees the full structure.
Since these buffers are now on the heap, we can adjust their size based
on the needs of the consumer. In particular, callers to
hashfd_throughput() are expecting to report progress indicators as the
buffer flushes. These callers would prefer the smaller 8k buffer to
avoid large delays between updates, especially for users with slower
networks. When the progress indicator is not used, the larger buffer is
preferrable.
By adding a new trace2 region in the chunk-format API, we can see that
the writing portion of 'git multi-pack-index write' lowers from ~1.49s
to ~1.47s on a Linux machine. These effects may be more pronounced or
diminished on other filesystems. The end-to-end timing is too noisy to
have a definitive change either way.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jonathan Nieder [Wed, 19 May 2021 01:52:56 +0000 (18:52 -0700)]
xsize_t: avoid implementation defined behavior when len < 0
The xsize_t helper aims to safely convert an off_t to a size_t,
erroring out when a file offset is too large to fit into a memory
address. It does this by using two casts:
size_t size = (size_t) len;
if (len != (off_t) size)
... error out ...
On a platform with sizeof(size_t) < sizeof(off_t), this check is safe
and correct. The first cast truncates to a size_t by finding the
remainder modulo SIZE_MAX+1 (see C99 section 6.3.1.3 Signed and
unsigned integers) and the second promotes to an off_t, meaning the
result is true if and only if len is representable as a size_t.
On other platforms, this two-casts strategy still works well (always
succeeds) for len >= 0. But for len < 0, when the first cast succeeds
and produces SIZE_MAX + 1 + len, the resulting value is too large to
be represented as an off_t, so the second cast produces implementation
defined behavior. In practice, it is likely to produce a result of
true despite len not being representable as size_t.
Simplify by replacing with a more straightforward check: compare len
to the relevant bounds and then cast it. (To avoid a -Wsign-compare
warning, after checking that len >= 0, we explicitly convert to a
sufficiently-large unsigned type before comparing to SIZE_MAX.)
In practice, this is not likely to come up since typical callers use
nonnegative len. Still, it's helpful to handle this case to make the
behavior easy to reason about.
Historical note: the original bounds-checking in 46be82dfd0 (xsize_t:
check whether we lose bits, 2010-07-28) did not produce this
implementation-defined behavior, though it still did not handle
negative offsets. It was not until 73560c793a (git-compat-util.h:
xsize_t() - avoid -Wsign-compare warnings, 2017-09-21) introduced the
double cast that the implementation-defined behavior was triggered.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
That commit does fix the situation it intended to (avoiding Negotiate
even when the credentials were provided in the URL), but it creates a
more serious regression: we now never hit the conditional for "we had a
username and password, tried them, but the server still gave us a 401".
That has two bad effects:
1. we never call credential_reject(), and thus a bogus credential
stored by a helper will live on forever
2. we never return HTTP_NOAUTH, so the error message the user gets is
"The requested URL returned error: 401", instead of "Authentication
failed".
Doing this correctly seems non-trivial, as we don't know whether the
Negotiate auth was a problem. Since this is a regression in the upcoming
v2.23.0 release (for which we're in -rc0), let's revert for now and work
on a fix separately.
(Note that this isn't a pure revert; the previous commit added a test
showing the regression, so we can now flip it to expect_success).
Reported-by: Ben Humphreys <behumphreys@atlassian.com> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Tue, 18 May 2021 06:27:36 +0000 (02:27 -0400)]
t5551: test http interaction with credential helpers
We test authentication with http, and we independently test that
credential helpers work, but we don't have any tests that cover the
two features working together. Let's add two:
1. Make sure that a successful request asks the helper to save the
credential. This works as expected.
2. Make sure that a failed request asks the helper to forget the
credential. This is marked as expect_failure, as it was recently
regressed by 1b0d9545bb (remote-curl: fall back to basic auth if
Negotiate fails, 2021-03-22). The symptom here is that the second
request should prompt the user, but doesn't.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Tue, 18 May 2021 00:53:37 +0000 (09:53 +0900)]
revisions(7): clarify that most commands take a single revision range
Sometimes new people are confused by how a revision "range" works,
in that it is not a random collection of commits but a set of
commits that are all connected to each other, and most Git commands
work on a single such "range".
Derrick Stolee [Mon, 17 May 2021 12:24:49 +0000 (12:24 +0000)]
hashfile: use write_in_full()
The flush() logic in csum-file.c was introduced originally by c38138c
(git-pack-objects: write the pack files with a SHA1 csum, 2005-06-26)
and a portion of the logic performs similar utility to write_in_full()
in wrapper.c. The history of write_in_full() is full of moves and
renames, but was originally introduced by 7230e6d (Add write_or_die(), a
helper function, 2006-08-21).
The point of these sections of code are to flush a write buffer using
xwrite() and report errors in the case of disk space issues or other
generic input/output errors. The logic in flush() can interpret the
output of write_in_full() to provide the correct error messages to
users.
The logic in the hashfile API has an additional set of logic to augment
the progress indicator between calls to xwrite(). This was introduced by 2a128d6 (add throughput display to git-push, 2007-10-30). It seems that
since the hashfile's buffer is only 8KB, these additional progress
indicators might not be incredibly necessary. Instead, update the
progress only when write_in_full() complete.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee [Mon, 17 May 2021 12:22:17 +0000 (12:22 +0000)]
sparse-index: fix uninitialized jump
While testing the sparse-index, I verified a test with --valgrind and it
complained about an uninitialized value being used in a jump in the
path_matches_pattern_list() method. The line was this one:
if (*dtype == DT_UNKNOWN)
In the call stack, the culprit was the initialization of the dtype
variable in convert_to_sparse_rec().
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Matheus Tavares [Mon, 17 May 2021 19:49:03 +0000 (16:49 -0300)]
parallel-checkout: send the new object_id algo field to the workers
An object_id storing a SHA-1 name has some unused bytes at the end of
the hash array. Since these bytes are not used, they are usually not
initialized to any value either. However, at
parallel_checkout.c:send_one_item() the object_id of a cache entry is
copied into a buffer which is later sent to a checkout worker through a
pipe write(). This makes Valgrind complain about passing uninitialized
bytes to a syscall. The worker won't use these uninitialized bytes
either, but the warning could confuse someone trying to debug this code;
So instead of using oidcpy(), send_one_item() uses hashcpy() to only
copy the used/initialized bytes of the object_id, and leave the
remaining part with zeros.
However, since cf0983213c ("hash: add an algo member to struct
object_id", 2021-04-26), using hashcpy() is no longer sufficient here as
it won't copy the new algo field from the object_id. Let's add and use a
new function which meets both our requirements of copying all the
important object_id data while still avoiding the uninitialized bytes,
by padding the end of the hash array in the destination object_id. With
this change, we also no longer need the destination buffer from
send_one_item() to be initialized with zeros, so let's switch from
xcalloc() to xmalloc() to make this clear.
Signed-off-by: Matheus Tavares <matheus.bernardino@usp.br> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Todd Zullinger [Mon, 17 May 2021 15:12:22 +0000 (11:12 -0400)]
t7500: remove non-existant C_LOCALE_OUTPUT prereq
The C_LOCALE_OUTPUT prerequisite was removed in b1e079807b (tests:
remove last uses of C_LOCALE_OUTPUT, 2021-02-11), where Ævar noted:
I'm not leaving the prerequisite itself in place for in-flight changes
as there currently are none that introduce new tests that rely on it,
and because C_LOCALE_OUTPUT is currently a noop on the master branch
we likely won't have any new submissions that use it.
One more use of C_LOCALE_OUTPUT did creep in with 3d1bda6b5b (t7500: add
tests for --fixup=[amend|reword] options, 2021-03-15). This causes a
number of the tests to be skipped by default:
ok 35 # SKIP --fixup=reword: incompatible with --all (missing C_LOCALE_OUTPUT)
ok 36 # SKIP --fixup=reword: incompatible with --include (missing C_LOCALE_OUTPUT)
ok 37 # SKIP --fixup=reword: incompatible with --only (missing C_LOCALE_OUTPUT)
ok 38 # SKIP --fixup=reword: incompatible with --interactive (missing C_LOCALE_OUTPUT)
ok 39 # SKIP --fixup=reword: incompatible with --patch (missing C_LOCALE_OUTPUT)
Remove the C_LOCALE_OUTPUT prerequisite from these tests so they are
not skipped.
Signed-off-by: Todd Zullinger <tmz@pobox.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Wolfgang Müller [Mon, 17 May 2021 08:02:42 +0000 (10:02 +0200)]
rev-parse: fix segfault with missing --path-format argument
Calling "git rev-parse --path-format" without an argument segfaults
instead of giving an error message. Commit fac60b8925 (rev-parse: add
option for absolute or relative path formatting, 2020-12-13) added the
argument parsing code but forgot to handle NULL.
Returning an error makes sense here because there is no default value we
could use. Add a test case to verify.
Signed-off-by: Wolfgang Müller <wolf@oriole.systems> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Daniel Santos [Mon, 15 Mar 2021 20:33:33 +0000 (20:33 +0000)]
l10n: pt_PT: add Portuguese translations part 2
* Eliminated 'Negation of emptiness' of 'nenhum' (not one/none)
* Eliminated 'Negation of emptiness' of 'nada' (nothing)
* Transformed 'Não' (No) into affirmative
* Some other translations
* Transforming 'não' (no) into affirmative
* From junção-de-3 to tri-junção
Signed-off-by: Daniel Santos <hello@brighterdan.com>
edef [Sun, 16 May 2021 15:07:19 +0000 (15:07 +0000)]
mailinfo: don't discard names under 3 characters
I sometimes receive patches from people with short mononyms, and in my
cultural environment these are not uncommon. To my dismay, git-am
currently discards their names, and replaces them with their email
addresses.
Gregory Anders [Fri, 14 May 2021 15:15:53 +0000 (09:15 -0600)]
git-send-email: add option to specify sendmail command
The sendemail.smtpServer configuration option and --smtp-server command
line option both support using a sendmail-like program to send emails by
specifying an absolute file path. However, this is not ideal for the
following reasons:
1. It overloads the meaning of smtpServer (now a program is being used
for the server?)
2. It doesn't allow for non-absolute paths, arguments, or arbitrary
scripting
Requiring an absolute path is bad for portability, as the same program
may be in different locations on different systems. If a user wishes to
pass arguments to their program, they have to use the smtpServerOption
option, which is cumbersome (as it must be repeated for each option) and
doesn't adhere to normal git conventions.
Introduce a new configuration option sendemail.sendmailCmd as well as a
command line option --sendmail-cmd that can be used to specify a command
(with or without arguments) or shell expression to run to send email.
The name of this option is consistent with --to-cmd and --cc-cmd. This
invocation honors the user's $PATH so that absolute paths are not
necessary. Arbitrary shell expressions are also supported, allowing
users to do basic scripting.
Give this option a higher precedence over --smtp-server and
sendemail.smtpServer, as the new interface is more flexible. For
backward compatibility, continue to support absolute paths in
--smtp-server and sendemail.smtpServer.
Signed-off-by: Gregory Anders <greg@gpanders.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Sun, 16 May 2021 12:05:24 +0000 (21:05 +0900)]
Merge branch 'dl/stash-show-untracked-fixup'
The code to handle options recently added to "git stash show"
around untracked part of the stash segfaulted when these options
were used on a stash entry that does not record untracked part.
* dl/stash-show-untracked-fixup:
stash show: fix segfault with --{include,only}-untracked
t3905: correct test title
Junio C Hamano [Sun, 16 May 2021 12:05:23 +0000 (21:05 +0900)]
Merge branch 'mt/parallel-checkout-part-3'
The final part of "parallel checkout".
* mt/parallel-checkout-part-3:
ci: run test round with parallel-checkout enabled
parallel-checkout: add tests related to .gitattributes
t0028: extract encoding helpers to lib-encoding.sh
parallel-checkout: add tests related to path collisions
parallel-checkout: add tests for basic operations
checkout-index: add parallel checkout support
builtin/checkout.c: complete parallel checkout support
make_transient_cache_entry(): optionally alloc from mem_pool
ZheNing Hu [Thu, 13 May 2021 15:15:38 +0000 (15:15 +0000)]
ref-filter: introduce enum atom_type
In the original ref-filter design, it will copy the parsed
atom's name and attributes to `used_atom[i].name` in the
atom's parsing step, and use it again for string matching
in the later specific ref attributes filling step. It use
a lot of string matching to determine which atom we need.
Introduce the enum "atom_type", each enum value is named
as `ATOM_*`, which is the index of each corresponding
valid_atom entry. In the first step of the atom parsing,
`used_atom.atom_type` will record corresponding enum value
from valid_atom entry index, and then in specific reference
attribute filling step, only need to compare the value of
the `used_atom[i].atom_type` to check the atom type.
Helped-by: Junio C Hamano <gitster@pobox.com> Helped-by: Christian Couder <christian.couder@gmail.com> Signed-off-by: ZheNing Hu <adlternative@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
ZheNing Hu [Thu, 13 May 2021 15:15:37 +0000 (15:15 +0000)]
ref-filter: add objectsize to used_atom
When the support for "objectsize:disk" was bolted onto the
existing support for "objectsize", it didn't follow the
usual pattern for handling "atomtype:modifier", which reads
the <modifier> part just once while parsing the format
string, and store the parsed result in the union in the
used_atom structure, so that the string form of it does not
have to be parsed over and over at runtime (e.g. in
grab_common_values()).
Add a new member `objectsize` to the union `used_atom.u`,
so that we can separate the check of <modifier> from the
check of <atomtype>, this will bring scalability to atom
`%(objectsize)`.
Signed-off-by: ZheNing Hu <adlternative@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King [Thu, 13 May 2021 06:25:53 +0000 (02:25 -0400)]
t: avoid sed-based chain-linting in some expensive cases
Commit 878f988350 (t/test-lib: teach --chain-lint to detect broken
&&-chains in subshells, 2018-07-11) introduced additional chain-lint
tests which add an extra "sed" pipeline to each test we run. This has a
measurable impact on runtime. Here are timings with and without a new
environment variable (added by this patch) that lets you disable just
the additional sed-based chain-lint tests:
Benchmark #1: GIT_TEST_CHAIN_LINT_HARDER=1 make test
Time (mean ± σ): 64.202 s ± 1.030 s [User: 622.469 s, System: 301.402 s]
Range (min … max): 61.571 s … 65.662 s 10 runs
Benchmark #2: GIT_TEST_CHAIN_LINT_HARDER=0 make test
Time (mean ± σ): 57.591 s ± 0.333 s [User: 529.368 s, System: 270.618 s]
Range (min … max): 57.143 s … 58.309 s 10 runs
Summary
'GIT_TEST_CHAIN_LINT_HARDER=0 make test' ran
1.11 ± 0.02 times faster than 'GIT_TEST_CHAIN_LINT_HARDER=1 make test'
Of course those extra lint checks are doing something useful, so paying
a few extra seconds (at least on Linux) isn't so bad (though note the
CPU time; we're bounded in our parallel run here by the slowest test, so
it really is ~120s of CPU improvement).
But we can observe that there are some test scripts where they produce a
much stronger effect, and provide less value. In t0027 and t3070 we run
a very large number of small tests, all driven by a series of
functions/loops which are filling in the test bodies. There we get much
less bang for our buck in terms of bug-finding versus CPU cost.
This patch introduces a mechanism for controlling when those extra
lint checks are run, at two levels:
- a user can ask to disable or to force-enable the checks by setting
GIT_TEST_CHAIN_LINT_HARDER
- if the user hasn't specified a preference, individual scripts can
disable the checks by setting GIT_TEST_CHAIN_LINT_HARDER_DEFAULT;
scripts which don't set that get the current behavior of enabling
them.
In addition, this patch flips the default for t0027 and t3070's
mass-generated sections to disable the extra checks. Here are the timing
results for t0027:
Benchmark #1: GIT_TEST_CHAIN_LINT_HARDER=1 ./t0027-auto-crlf.sh
Time (mean ± σ): 17.078 s ± 0.848 s [User: 14.878 s, System: 7.075 s]
Range (min … max): 15.952 s … 18.421 s 10 runs
Benchmark #2: GIT_TEST_CHAIN_LINT_HARDER=0 ./t0027-auto-crlf.sh
Time (mean ± σ): 9.063 s ± 0.759 s [User: 7.890 s, System: 3.362 s]
Range (min … max): 7.747 s … 10.619 s 10 runs
Benchmark #3: ./t0027-auto-crlf.sh
Time (mean ± σ): 9.186 s ± 0.881 s [User: 7.957 s, System: 3.427 s]
Range (min … max): 7.796 s … 10.498 s 10 runs
Summary
'GIT_TEST_CHAIN_LINT_HARDER=0 ./t0027-auto-crlf.sh' ran
1.01 ± 0.13 times faster than './t0027-auto-crlf.sh'
1.88 ± 0.18 times faster than 'GIT_TEST_CHAIN_LINT_HARDER=1 ./t0027-auto-crlf.sh'
We can see that disabling the checks for the whole script buys us an
almost 2x speedup. But the new default behavior, disabling them only for
the mass-generated part, gets us most of that speedup (but still leaves
the checks on for further manual tests people might write).
As a side note, I'd caution about comparing runtimes and CPU seconds
between this timing and the earlier "make test" one. In "make test",
we're running a lot of scripts in parallel, so the CPU is throttling
down (and thus a CPU second saved here would count for more during a
parallel run; the same work takes more CPU seconds there).
We get similar results for t3070:
Benchmark #1: GIT_TEST_CHAIN_LINT_HARDER=1 ./t3070-wildmatch.sh
Time (mean ± σ): 20.054 s ± 3.967 s [User: 16.003 s, System: 8.286 s]
Range (min … max): 11.891 s … 23.671 s 10 runs
Benchmark #2: GIT_TEST_CHAIN_LINT_HARDER=0 ./t3070-wildmatch.sh
Time (mean ± σ): 12.399 s ± 2.256 s [User: 7.542 s, System: 5.342 s]
Range (min … max): 9.606 s … 15.727 s 10 runs
Benchmark #3: ./t3070-wildmatch.sh
Time (mean ± σ): 10.726 s ± 3.476 s [User: 6.790 s, System: 4.365 s]
Range (min … max): 5.444 s … 15.376 s 10 runs
Summary
'./t3070-wildmatch.sh' ran
1.16 ± 0.43 times faster than 'GIT_TEST_CHAIN_LINT_HARDER=0 ./t3070-wildmatch.sh'
1.87 ± 0.71 times faster than 'GIT_TEST_CHAIN_LINT_HARDER=1 ./t3070-wildmatch.sh'
Again, we get almost a 2x speedup disabling these. In this case, there
are no tests not covered by the script's "default to disable" behavior,
so the second two benchmarks should be the same (and while they do
differ, you can see the variance is quite high but they're within one
standard deviation).
So it seems like for these two scripts, at least, disabling the extra
checks is a reasonable tradeoff. Sadly, the overall runtime of "make
test" on my system doesn't get much faster. But that's because we're
mostly limited by the cost of the single biggest test. Here are the
top-5 tests by wall-clock time from a parallel run, before my patch:
We dropped 12s from t0027, and t3070 dropped off our list entirely at
around 16s. In both cases we're bound by t9001, but its slowness is
due to the actual tests, so we'll have to deal with it in a different
way. But this reduces overall CPU, and means that dealing with t9001 (by
improving the speed of send-email or splitting it apart) will let us
reduce our overall runtime even on multi-core machines.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Thu, 13 May 2021 06:22:36 +0000 (06:22 +0000)]
git-prompt: work under set -u
Commit afda36dbf3 ("git-prompt: include sparsity state as well",
2020-06-21) added the use of some variables to control how to show
sparsity state in the git prompt, but implicitly assumed that undefined
variables would be treated as the empty string. This breaks users who
run under 'set -u'; fix the code to be more explicit.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Denton Liu [Wed, 12 May 2021 20:16:13 +0000 (13:16 -0700)]
stash show: fix segfault with --{include,only}-untracked
When `git stash show --include-untracked` or
`git stash show --only-untracked` is run on a stash that doesn't include
an untracked entry, a segfault occurs. This happens because we do not
check whether the untracked entry is actually present and just attempt
to blindly dereference it.
Ensure that the untracked entry is present before actually attempting to
dereference it.
Signed-off-by: Denton Liu <liu.denton@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many places in the code were doing
while ((d = readdir(dir)) != NULL) {
if (is_dot_or_dotdot(d->d_name))
continue;
...process d...
}
Introduce a readdir_skip_dot_and_dotdot() helper to make that a one-liner:
while ((d = readdir_skip_dot_and_dotdot(dir)) != NULL) {
...process d...
}
This helper particularly simplifies checks for empty directories.
Also use this helper in read_cached_dir() so that our statistics are
consistent across platforms. (In other words, read_cached_dir() should
have been using is_dot_or_dotdot() and skipping such entries, but did
not and left it to treat_path() to detect and mark such entries as
path_none.)
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Derrick Stolee [Wed, 12 May 2021 17:28:21 +0000 (17:28 +0000)]
dir: update stale description of treat_directory()
The documentation comment for treat_directory() was originally written
in 095952 (Teach directory traversal about subprojects, 2007-04-11)
which was before the 'struct dir_struct' split its bitfield of named
options into a 'flags' enum in 7c4c97c0 (Turn the flags in struct
dir_struct into a single variable, 2009-02-16). When those flags
changed, the comment became stale, since members like
'show_other_directories' transitioned into flags like
DIR_SHOW_OTHER_DIRECTORIES.
Update the comments for treat_directory() to use these flag names rather
than the old member names.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Wed, 12 May 2021 17:28:20 +0000 (17:28 +0000)]
dir: traverse into untracked directories if they may have ignored subfiles
A directory that is untracked does not imply that all files under it
should be categorized as untracked; in particular, if the caller is
interested in ignored files, many files or directories underneath the
untracked directory may be ignored. We previously partially handled
this right with DIR_SHOW_IGNORED_TOO, but missed DIR_SHOW_IGNORED. It
was not obvious, though, because the logic for untracked and excluded
files had been fused together making it harder to reason about. The
previous commit split that logic out, making it easier to notice that
DIR_SHOW_IGNORED was missing. Add it.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Wed, 12 May 2021 17:28:19 +0000 (17:28 +0000)]
dir: avoid unnecessary traversal into ignored directory
The show_other_directories case in treat_directory() tried to handle
both excludes and untracked files with the same logic, and mishandled
both the excludes and the untracked files in the process, in different
ways. Split that logic apart, and then focus on the logic for the
excludes; a subsequent commit will address the logic for untracked
files.
For show_other_directories, an excluded directory means that
every path underneath that directory will also be excluded. Given that
the calling code requested to just show directories when everything
under a directory had the same state (that's what the
"DIR_SHOW_OTHER_DIRECTORIES" flag means), we generally do not need to
traverse into such directories and can just immediately mark them as
ignored (i.e. as path_excluded). The only reason we cannot just
immediately return path_excluded is the DIR_HIDE_EMPTY_DIRECTORIES flag
and the possibility that the ignored directory is an empty directory.
The code previously treated DIR_SHOW_IGNORED_TOO in most cases as an
exception as well, which was wrong. It can sometimes reduce the number
of cases where we need to recurse (namely if
DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set), but should not be able
to increase the number of cases where we need to recurse. Fix the logic
accordingly.
Some sidenotes about possible confusion with dir.c:
* "ignored" often refers to an untracked ignore", i.e. a file which is
not tracked which matches one of the ignore/exclusion rules. But you
can also have a "tracked ignore", a tracked file that happens to match
one of the ignore/exclusion rules and which dir.c has to worry about
since "git ls-files -c -i" is supposed to list them.
* The dir code often uses "ignored" and "excluded" interchangeably,
which you need to keep in mind while reading the code.
* "exclude" is used multiple ways in the code:
* As noted above, "exclude" is often a synonym for "ignored".
* The logic for parsing .gitignore files was re-used in
.git/info/sparse-checkout, except there it is used to mark paths that
the user wants to *keep*. This was mostly addressed by commit 65edd96aec ("treewide: rename 'exclude' methods to 'pattern'",
2019-09-03), but every once in a while you'll find a comment about
"exclude" referring to these patterns that might in fact be in use
by the sparse-checkout machinery for inclusion rules.
* The word "EXCLUDE" is also used for pathspec negation, as in
(pathspec->items[3].magic & PATHSPEC_EXCLUDE)
Thus if a user had a .gitignore file containing
*~
*.log
!settings.log
And then ran
git add -- 'settings.*' ':^settings.log'
Then :^settings.log is a pathspec negation making settings.log not
be requested to be added even though all other settings.* files are
being added. Also, !settings.log in the gitignore file is a negative
exclude pattern meaning that settings.log is normally a file we
want to track even though all other *.log files are ignored.
Sometimes it feels like dir.c needs its own glossary with its many
definitions, including the multiply-defined terms.
Reported-by: Jason Gore <Jason.Gore@microsoft.com> Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the last commit, we added a testcase showing that the directory
traversal machinery sometimes traverses into directories unnecessarily.
Here we show that there are cases where it does the opposite: it does
not traverse into directories, despite those directories having
important files that need to be flagged.
Add a testcase showing that `git ls-files -o -i --directory` can omit
some of the files it should be listing, and another showing that `git
clean -fX` can fail to clean out some of the expected files.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Wed, 12 May 2021 17:28:17 +0000 (17:28 +0000)]
t7300: add testcase showing unnecessary traversal into ignored directory
The PNPM package manager is apparently creating deeply nested (but
ignored) directory structures; traversing them is costly
performance-wise, unnecessary, and in some cases is even throwing
warnings/errors because the paths are too long to handle on various
platforms. Add a testcase that checks for such unnecessary directory
traversal.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Wed, 12 May 2021 17:28:16 +0000 (17:28 +0000)]
ls-files: error out on -i unless -o or -c are specified
ls-files --ignored can be used together with either --others or
--cached. After being perplexed for a bit and digging in to the code, I
assumed that ls-files -i was just broken and not printing anything and
I had a nice patch ready to submit when I finally realized that -i can be
used with --cached to find tracked ignores.
While that was a mistake on my part, and a careful reading of the
documentation could have made this more clear, I suspect this is an
error others are likely to make as well. In fact, of two uses in our
testsuite, I believe one of the two did make this error. In t1306.13,
there are NO tracked files, and all the excludes built up and used in
that test and in previous tests thus have to be about untracked files.
However, since they were looking for an empty result, the mistake went
unnoticed as their erroneous command also just happened to give an empty
answer.
-i will most the time be used with -o, which would suggest we could just
make -i imply -o in the absence of either a -o or -c, but that would be
a backward incompatible break. Instead, let's just flag -i without
either a -o or -c as an error, and update the two relevant testcases to
specify their intent.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Wed, 12 May 2021 17:28:15 +0000 (17:28 +0000)]
dir: report number of visited directories and paths with trace2
Provide more statistics in trace2 output that include the number of
directories and total paths visited by the directory traversal logic.
Subsequent patches will take advantage of this to ensure we do not
unnecessarily traverse into ignored directories.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 07d90eadb50 (Makefile: add Perl runtime prefix support,
2018-04-10) PERL_DEFINES has been a simply-expanded variable, let's
make it recursively expanded instead.
This change doesn't matter for the correctness of the logic. Whether
we used simply-expanded or recursively expanded didn't change what we
wrote out in GIT-PERL-DEFINES, but being consistent with other rules
makes this easier to understand.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Eric Wong [Tue, 11 May 2021 10:37:30 +0000 (10:37 +0000)]
remote-curl: fix clone on sha256 repos
The remote-https process needs to update it's own instance of
`the_repository' when it sees an HTTP(S) remote is using sha256.
Without this, parse_oid_hex() fails to handle sha256 OIDs when
it's eventually called by parse_fetch().
ZheNing Hu [Tue, 11 May 2021 15:35:20 +0000 (15:35 +0000)]
ref-filter: fix read invalid union member bug
used_atom.u is an union, and it has different members depending on
what atom the auxiliary data the union part of the "struct
used_atom" wants to record. At most only one of the members can be
valid at any one time. Since the code checks u.remote_ref without
even making sure if the atom is "push" or "push:" (which are only
two cases that u.remote_ref.push becomes valid), but u.remote_ref
shares the same storage for other members of the union, the check
was reading from an invalid member, which was the bug.
Modify the condition here to check whether the atom name
equals to "push" or starts with "push:", to avoid reading the
value of invalid member of the union.
Signed-off-by: ZheNing Hu <adlternative@gmail.com>
[jc: further test fixes] Signed-off-by: Junio C Hamano <gitster@pobox.com>
Lénaïc Huard [Mon, 10 May 2021 19:59:09 +0000 (21:59 +0200)]
maintenance: fix two memory leaks
Fixes two memory leaks when running `git maintenance start` or `git
maintenance stop` in `update_background_schedule`:
$ valgrind --leak-check=full ~/git/bin/git maintenance start
==76584== Memcheck, a memory error detector
==76584== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==76584== Using Valgrind-3.16.1 and LibVEX; rerun with -h for copyright info
==76584== Command: /home/lenaic/git/bin/git maintenance start
==76584==
==76584==
==76584== HEAP SUMMARY:
==76584== in use at exit: 34,880 bytes in 252 blocks
==76584== total heap usage: 820 allocs, 568 frees, 146,414 bytes allocated
==76584==
==76584== 65 bytes in 1 blocks are definitely lost in loss record 17 of 39
==76584== at 0x483E6AF: malloc (vg_replace_malloc.c:306)
==76584== by 0x3DC39C: xrealloc (wrapper.c:126)
==76584== by 0x3992CC: strbuf_grow (strbuf.c:98)
==76584== by 0x39A473: strbuf_vaddf (strbuf.c:392)
==76584== by 0x39BC54: xstrvfmt (strbuf.c:979)
==76584== by 0x39BD2C: xstrfmt (strbuf.c:989)
==76584== by 0x18451B: update_background_schedule (gc.c:1977)
==76584== by 0x1846F6: maintenance_start (gc.c:2011)
==76584== by 0x1847B4: cmd_maintenance (gc.c:2030)
==76584== by 0x127A2E: run_builtin (git.c:453)
==76584== by 0x127E81: handle_builtin (git.c:704)
==76584== by 0x128142: run_argv (git.c:771)
==76584==
==76584== 240 bytes in 1 blocks are definitely lost in loss record 29 of 39
==76584== at 0x4840D7B: realloc (vg_replace_malloc.c:834)
==76584== by 0x491CE5D: getdelim (in /usr/lib/libc-2.33.so)
==76584== by 0x39ADD7: strbuf_getwholeline (strbuf.c:635)
==76584== by 0x39AF31: strbuf_getdelim (strbuf.c:706)
==76584== by 0x39B064: strbuf_getline_lf (strbuf.c:727)
==76584== by 0x184273: crontab_update_schedule (gc.c:1919)
==76584== by 0x184678: update_background_schedule (gc.c:1997)
==76584== by 0x1846F6: maintenance_start (gc.c:2011)
==76584== by 0x1847B4: cmd_maintenance (gc.c:2030)
==76584== by 0x127A2E: run_builtin (git.c:453)
==76584== by 0x127E81: handle_builtin (git.c:704)
==76584== by 0x128142: run_argv (git.c:771)
==76584==
==76584== LEAK SUMMARY:
==76584== definitely lost: 305 bytes in 2 blocks
==76584== indirectly lost: 0 bytes in 0 blocks
==76584== possibly lost: 0 bytes in 0 blocks
==76584== still reachable: 34,575 bytes in 250 blocks
==76584== suppressed: 0 bytes in 0 blocks
==76584== Reachable blocks (those to which a pointer was found) are not shown.
==76584== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==76584==
==76584== For lists of detected and suppressed errors, rerun with: -s
==76584== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 0 from 0)
Signed-off-by: Lénaïc Huard <lenaic@lhuard.fr> Acked-by: Derrick Stolee <stolee@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Tue, 11 May 2021 06:27:23 +0000 (15:27 +0900)]
Merge branch 'zh/trailer-cmd'
The way the command line specified by the trailer.<token>.command
configuration variable receives the end-user supplied value was
both error prone and misleading. An alternative to achieve the
same goal in a safer and more intuitive way has been added, as
the trailer.<token>.cmd configuration variable, to replace it.
* zh/trailer-cmd:
trailer: add new .cmd config option
docs: correct descript of trailer.<token>.command
Junio C Hamano [Tue, 11 May 2021 06:27:23 +0000 (15:27 +0900)]
Merge branch 'jk/symlinked-dotgitx-cleanup'
Various test and documentation updates about .gitsomething paths
that are symlinks.
* jk/symlinked-dotgitx-cleanup:
docs: document symlink restrictions for dot-files
fsck: warn about symlinked dotfiles we'll open with O_NOFOLLOW
t0060: test ntfs/hfs-obscured dotfiles
t7450: test .gitmodules symlink matching against obscured names
t7450: test verify_path() handling of gitmodules
t7415: rename to expand scope
fsck_tree(): wrap some long lines
fsck_tree(): fix shadowed variable
t7415: remove out-dated comment about translation
Options to "git pack-objects" that take numeric values like
--window and --depth should not accept negative values; the input
validation has been tightened.
* jk/pack-objects-negative-options-fix:
pack-objects: clamp negative depth to 0
t5316: check behavior of pack-objects --depth=0
pack-objects: clamp negative window size to 0
t5300: check that we produced expected number of deltas
t5300: modernize basic tests
Will Chandler [Sat, 8 May 2021 05:00:43 +0000 (01:00 -0400)]
refs: cleanup directories when deleting packed ref
When deleting a packed ref via 'update-ref -d', a lockfile is made in
the directory that would contain the loose copy of that ref, creating
any directories in the ref's path that do not exist. When the
transaction completes, the lockfile is deleted, but any empty parent
directories made when creating the lockfile are left in place. These
empty directories are not removed by 'pack-refs' or other housekeeping
tasks and will accumulate over time.
When deleting a loose ref, we remove all empty parent directories at the
end of the transaction.
This commit applies the parent directory cleanup logic used when
deleting loose refs to packed refs as well.
Signed-off-by: Will Chandler <wfc@wfchandler.org> Reviewed-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change a subshell added in a preceding commit to instead use a new
"-C" option to "check_describe". The idiom for this is copied as-is
from the "test_commit" function in test-lib-functions.sh
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix a nested invocation of "test_expect_success", the
"check_describe()" function is a wrapper for calling
test_expect_success, and therefore needs to be called outside the body
of another "test_expect_success".
The two tests added in 30b1c7ad9d6 (describe: don't abort too early
when searching tags, 2020-02-26) were not testing for anything due to
this logic error. Without this fix reverting the C code changes in
that commit still has all tests passing, with this fix we're actually
testing the "describe" output. This is because "test_expect_success"
calls "test_finish_", whose last statement happens to be true.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
describe tests: don't rely on err.actual from "check_describe"
Convert the one test that relied on the "err.actual" file produced by
check_describe() to instead do its own check of "git describe"
output.
This means that the two tests won't have an inter-dependency (e.g. if
the earlier test is skipped).
An earlier version of this patch instead asserted that no other test
had any output on stderr. We're not doing that here out of fear that
"gc --auto" or another future change to "git describe" will cause it
to legitimately emit output on stderr unexpectedly[1].
I'd think that inverting the test added in 3291fe4072e (Add
git-describe test for "verify annotated tag names on output",
2008-03-03) to make checking that we don't have warnings the rule
rather than the exception would be the sort of thing the describe
tests should be catching, but for now let's leave it as it is.
Change the glob matching via a "case" statement to a "test_cmp" after
we've stripped out the hash-specific g<hash-abbrev>
suffix. 5312ab11fbf (Add describe test., 2007-01-13).
This means that we can use test_cmp to compare the output. I could
omit the "-8" change of e.g. "A-*" to "A-8-gHASH", but I think it
makes sense to test that here explicitly. It means you need to add new
tests to the bottom of the file, but that's not a burden in this case.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
describe tests: improve test for --work-tree & --dirty
Improve tests added in 9f67d2e8279 (Teach "git describe" --dirty
option, 2009-10-21) and 2ed5c8e174d (describe: setup working tree for
--dirty, 2019-02-03) so that they make sense in combination with each
other.
The "check_describe" being removed here was the earlier test, we then
later added these --work-tree tests which really just wanted to check
if we got the exact same output from "describe", but the test wasn't
structured to test for that.
Let's change it to do that, which both improves test coverage and
makes it more obvious what's going on here.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
xdiff-interface: replace discard_hunk_line() with a flag
Remove the dummy discard_hunk_line() function added in 3b40a090fd4 (diff: avoid generating unused hunk header lines,
2018-11-02) in favor of having a new XDL_EMIT_NO_HUNK_HDR flag, for
use along with the two existing and similar XDL_EMIT_* flags.
Unlike the recently amended xdiff_emit_line_fn interface which'll be
called in a loop in xdl_emit_diff(), the hunk header is only emitted
once.
It makes more sense to pass this as a flag than provide a dummy
callback because that function may be able to skip doing certain work
if it knows the caller is doing nothing with the hunk header.
It would be possible to do so in the case of -U0 now, but the benefit
of doing so is so small that I haven't bothered. But this leaves the
door open to that, and more importantly makes the API use more
intuitive.
The reason we're putting a flag in the gap between 1<<0 and 1<<2 is
that the old 1<<1 flag was removed in 907681e940d (xdiff: drop
XDL_EMIT_COMMON, 2016-02-23) without re-ordering the remaining flags.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
xdiff users: use designated initializers for out_line
Amend the code added in 611e42a5980 (xdiff: provide a separate emit
callback for hunks, 2018-11-02) to be more readable by using
designated initializers.
This changes "priv" in rerere.c to be initialized to NULL as we did in
merge-tree.c. That's not needed as we'll only use it if the callback
is defined, but being consistent here is better and less verbose.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of special-casing creations and deletions let's just generate
a diff for them.
This logic of not running a diff under -G if we don't have both sides
dates back to the original implementation of -S in 52e9578985f ([PATCH] Introducing software archaeologist's tool
"pickaxe"., 2005-05-21).
In the case of -S we were not working with the xdiff interface and
needed to do this, but when -G was implemented in f506b8e8b5f (git
log/diff: add -G<regexp> that greps in the patch text, 2010-08-23)
this logic was diligently copied over.
But as the performance test added earlier in this series shows, this
does not make much of a difference. With:
time GIT_TEST_LONG= GIT_PERF_REPEAT_COUNT=10 GIT_PERF_MAKE_OPTS='-j8 CFLAGS=-O3' ./run origin/next HEAD~ HEAD -- p4209-pickaxe.sh
With the HEAD~ commit being the preceding "pickaxe -G: terminate early
on matching lines" we get these results. Note that it's only the -G
codepaths that are relevant to this change:
This small simplification really doesn't buy us much now, but I've got
plans to both convert the pickaxe code to using a PCREv2 backend[1]
and to implement additional pickaxe modes to do custom searches
through the diff[2]. Always having the diff available under -G is
going to help to simplify both of those changes.
Solve a long-standing item for "git log -Grx" of us e.g. finding "+
str" in the diff context and noting that we had a "hit", but xdiff
diligently continuing to generate and spew the rest of the diff at
us. This makes use of a new "early return" xdiff interface added by
preceding commits.
The TODO item (or, the NEEDSWORK comment) has been there since "git
log -G" was implemented. See f506b8e8b5f (git log/diff: add -G<regexp>
that greps in the patch text, 2010-08-23).
But now with the support added in the preceding changes to the
xdiff-interface we can return early. Let's assert the behavior of that
new early-return xdiff-interface by having a BUG() call here to die if
it ever starts handing us needless work again.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
xdiff-interface: allow early return from xdiff_emit_line_fn
Finish the change started in the preceding commit and allow an early
return from "xdiff_emit_line_fn" callbacks, this will allows
diffcore-pickaxe.c to save itself redundant work.
Our xdiff interface also had the limitation of not being able to abort
early since the beginning, see d9ea73e0564 (combine-diff: refactor
built-in xdiff interface., 2006-04-05). Although at that time
"xdiff_emit_line_fn" was called "xdiff_emit_consume_fn", and
"xdiff_emit_hunk_fn" didn't exist yet.
There was some work in this area of xdiff-interface.[ch] recently with 3b40a090fd4 (diff: avoid generating unused hunk header lines,
2018-11-02) and 7c61e25fbf1 (diff: use hunk callback for word-diff,
2018-11-02).
In combination those two changes allow us to not do any work on the
hunks and diff at all, but didn't change the status quo with regards
to consumers that e.g. want the diff lines, but might want to abort
early.
Whereas now we can abort e.g. on the first "-line" of a 1000 line diff
if that's all we needed.
This interface is rather scary as noted in the comment to
xdiff-interface.h being added here, as noted there a future change
could add more exit codes, and hack xdl_emit_diff() and friends to
ignore or skip things more selectively as a result.
I did not see an inherent reason for why xdl_emit_{diffrec,record}()
could not be changed to ferry the "xdiff_emit_line_fn" error code
upwards instead of returning -1 on all "ret < 0".
But doing so would require corresponding changes in xdl_emit_diff(),
xdl_diff(). I didn't see any issue with narrowly doing that to
accomplish what I needed here, but it would leave xdiff's own return
values in an inconsistent state.
Instead I've left it at returning a more conventional (for git's own
codebase) 1 for an early return, and translating it (or rather, all
non-zero) to -1 for xdiff's consumption.
The reason for most of the "stop" complexity in xdiff_outf() is
because we want to be able to abort early, but do so in a way that
doesn't skip the appropriate strbuf_reset() invocations.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
xdiff-interface: prepare for allowing early return
Change the function prototype of xdiff_emit_line_fn to return an "int"
instead of "void". Change all of those functions to "return 0",
nothing checks those return values yet, and no behavior is being
changed.
In subsequent commits the interface will be changed to allow early
return via this new return value.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the "log -S<pat>" switch counts occurrences of <pat> on the
pre-image and post-image of a change. As soon as we know we had e.g. 1
before and 2 now we can stop, we don't need to keep counting past 2.
With this change a diff between A and B may have different performance
characteristics than between B and A. That's OK in this case, since
we'll emit the same output, and the effect is to make one of them
better.
I'm picking a check of "one" first on the assumption that it's a more
common case to have files grow over time than not.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
pickaxe -S: support content with NULs under --pickaxe-regex
Fix a bug in the matching routine powering -S<rx> --pickaxe-regex so
that we won't abort early on content that has NULs in it.
We've had a hard requirement on REG_STARTEND since 2f8952250a8 (regex:
add regexec_buf() that can work on a non NUL-terminated string,
2016-09-21), but this sanity check dates back to d01d8c67828 (Support
for pickaxe matching regular expressions, 2006-03-29).
It wasn't needed anymore, and as the now-passing test shows, actively
getting in our way. Since we always require REG_STARTEND support we do
not need to stop at NULs. If we are dealing with a haystack with NUL
in it. The needle may be behind that NUL.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
pickaxe: assert that we must have a needle under -G or -S
Assert early in diffcore_pickaxe() that we've got a needle to work
with under -G and -S.
This code is redundant to the check -G and -S get from
parse-options.c's get_arg(), which I'm adding a test for.
This check dates back to e1b161161d (diffcore-pickaxe: fix infinite
loop on zero-length needle, 2007-01-25) when "git log -S" could send
this code into an infinite loop.
It was then later refactored in 8fa4b09fb1 (pickaxe: hoist empty
needle check, 2012-10-28) into its current form, but it seemingly
wasn't noticed that in the meantime a move to the parse-options.c API
in dea007fb4c (diff: parse separate options like -S foo, 2010-08-05)
had made it redundant.
Let's retain some of the paranoia here with a BUG(), but there's no
need to be checking this in the pickaxe_match() inner loop.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
pickaxe: refactor function selection in diffcore-pickaxe()
It's hard to read this codepath at a glance and reason about exactly
what combination of -G and -S will compile either regexes or kwset,
and whether we'll then dispatch to "diff_grep" or "has_changes".
Then in the "--find-object" case we aren't using the callback
function, but were previously passing down "has_changes".
Refactor this code to exhaustively check "opts", it's now more obvious
what callback function (or none) we want under what mode.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a test for the -G and -S pickaxe options and related options.
This test supports being run with GIT_TEST_LONG=1 to adjust the limit
on the number of commits from 1k to 10k. The 1k limit seems to hit a
good spot on git.git
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
projects / thirdparty / git.git / log
