Using "1~5" isn't portable. Nobody seems to have noticed, since perhaps
people don't tend to run the perf suite on more exotic platforms. Still,
it's better to set a good example.
We can use:
perl -ne 'print if $. % 5 == 1'
instead. But we can further observe that perl does a good job of the
other parts of this pipeline, and fold the whole thing together.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Right now, the test suite can be run with 'GIT_TEST_WRITE_REV_INDEX=1'
in the environment, which causes all operations which write a pack to
also write a .rev file.
To prepare for when that eventually becomes the default, we should
continue to test the in-memory reverse index, too, in order to avoid
losing existing coverage. Unfortunately, explicit existing coverage is
rather sparse, so only a basic test is added that compares the result of
git rev-list --objects --no-object-names --all |
git cat-file --batch-check='%(objectsize:disk) %(objectname)'
with and without an on-disk reverse index.
Suggested-by: Jeff King <peff@peff.net>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we expand a user-format, we try to avoid work that isn't necessary
for the output. For instance, we don't bother parsing the commit header
until we know we need the author, subject, etc.
But we do always load the commit object's contents from disk, even if
the format doesn't require it (e.g., just "%H"). Traditionally this
didn't matter much, because we'd have loaded it as part of the traversal
anyway, and we'd typically have those bytes attached to the commit
struct (or these days, cached in a commit-slab).
But when we have a commit-graph, we might easily get to the point of
pretty-printing a commit without ever having looked at the actual object
contents. We should push off that load (and reencoding) until we're
certain that it's needed.
I think the results of p4205 show the advantage pretty clearly (we serve
parent and tree oids out of the commit struct itself, so they benefit as
well):
# using git.git as the test repo
Test HEAD^ HEAD
----------------------------------------------------------------------
4205.1: log with %H 0.40(0.39+0.01) 0.03(0.02+0.01) -92.5%
4205.2: log with %h 0.45(0.44+0.01) 0.09(0.09+0.00) -80.0%
4205.3: log with %T 0.40(0.39+0.00) 0.04(0.04+0.00) -90.0%
4205.4: log with %t 0.46(0.46+0.00) 0.09(0.08+0.01) -80.4%
4205.5: log with %P 0.39(0.39+0.00) 0.03(0.03+0.00) -92.3%
4205.6: log with %p 0.46(0.46+0.00) 0.10(0.09+0.00) -78.3%
4205.7: log with %h-%h-%h 0.52(0.51+0.01) 0.15(0.14+0.00) -71.2%
4205.8: log with %an-%ae-%s 0.42(0.41+0.00) 0.42(0.41+0.01) +0.0%
# using linux.git as the test repo
Test HEAD^ HEAD
----------------------------------------------------------------------
4205.1: log with %H 7.12(6.97+0.14) 0.76(0.65+0.11) -89.3%
4205.2: log with %h 7.35(7.19+0.16) 1.30(1.19+0.11) -82.3%
4205.3: log with %T 7.58(7.42+0.15) 1.02(0.94+0.08) -86.5%
4205.4: log with %t 8.05(7.89+0.15) 1.55(1.41+0.13) -80.7%
4205.5: log with %P 7.12(7.01+0.10) 0.76(0.69+0.07) -89.3%
4205.6: log with %p 7.38(7.27+0.10) 1.32(1.20+0.12) -82.1%
4205.7: log with %h-%h-%h 7.81(7.67+0.13) 1.79(1.67+0.12) -77.1%
4205.8: log with %an-%ae-%s 7.90(7.74+0.15) 7.81(7.66+0.15) -1.1%
I added the final test to show where we don't improve (the 1% there is
just lucky noise), but also as a regression test to make sure we're not
doing anything stupid like loading the commit multiple times when there
are several placeholders that need it.
Reported-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In 6e98de72c0 (sequencer (rebase -i): add support for the 'fixup' and
'squash' commands, 2017-01-02), this developer introduced a change of
behavior by mistake: when encountering a `fixup!` commit (or multiple
`fixup!` commits) without any `squash!` commit thrown in, the final `git
commit` was invoked with `--cleanup=strip`. Prior to that commit, the
commit command had been called without that `--cleanup` option.
Since we explicitly read the original commit message from a file in that
case, there is really no sense in forcing that clean-up.
We actually need to actively suppress that clean-up lest a configured
`commit.cleanup` may interfere with what we want to do: leave the commit
message unchanged.
Reported-by: Vojtěch Knyttl <vojtech@knyt.tl>
Helped-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we use the sub-test helpers, we end up defining one shell snippet
inside another shell snippet. So if we use single-quotes for the outer
snippet, we have to use double-quotes within the inner snippet (it's
included as here-doc within the outer snippet, but using a single quote
would end the outer snippet early). Or vice versa we can use double
quotes for the outer snippet, but then single quotes in the inner.
We have some of each in the script, and neither is wrong. But it would
be nice to be consistent unless there is a good reason not to. Using
single quotes for the outer script is preferable, because it requires
less metacharacter quoting overall. For example, in:
test_expect_success 'outer' '
run_sub_test_lib_test ... <<-\EOF
echo $foo &&
test_expect_success "inner" "
echo \$bar
"
EOF
'
we need only quote inside "inner", but not inside "outer" or the
here-doc. Whereas if we flip them, we have to quote in both places:
test_expect_success 'outer' "
run_sub_test_lib_test ... <<-\EOF
echo \$foo &&
test_expect_success 'inner' '
echo \$bar
'
EOF
"
The exception is when we need a literal single-quote in an expected
output here-doc. There we can either use outer double-quotes, or just
use ${SQ} within the doc. I chose the latter for consistency (within
this test, but also with other test scripts that face the same problem).
There is one other interesting case, which is some tests that do:
test_expect_success ... "
do_something --run='"'!3'"'
"
This is rather confusing to read, but is correct. The outer script sees
'!3' in single-quotes, as does the eval'd snippet. This is perhaps being
overly cautious. In many interactive shells, an exclamation triggers
history expansion even inside double quotes, but that is not generally
true in non-interactive shells.
There's some conflicting information here. Commit 784ce03d55 (t4216:
avoid unnecessary subshell in test_bloom_filters_not_used, 2020-05-19)
reports it as a problem with OpenBSD 6.7's /bin/sh. However, we have
many instances in this script of prereqs like !LAZY_TRUE, which haven't
been a problem. I left them un-escaped here to test out this theory.
It's much nicer if we can not worry about this as a portability issue,
so it's worth knowing.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our check of test_when_finished is done directly in the main script, and
if we failed to clean, we complain and exit immediately. It's nicer to
signal a test failure here, for a few reasons:
- this gives better output to the user when run under a TAP harness
like "prove"
- constency; it's the only test left in the file that behaves this way
- half of its "if" conditional is nonsense anyway; it picked up a
reference to GIT_TEST_FAIL_PREREQS_INTERNAL in dfe1a17df9 (tests:
add a special setup where prerequisites fail, 2019-05-13) along with
its neighbors, even though it has nothing to do with that flag
We could actually do this without a sub-test at all, and just put our
two tests (one to do cleanup, and one to check that it happened) in the
main script. But doing it in a subtest is conceptually cleaner (from the
perspective of the main test script, we are checking only one thing),
and it remains consistent with the "cleanup when failing" test directly
after it, which has to happen in a sub-test (to avoid the main script
complaining of the failed test).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We test the behavior of prerequisites in t0000 by setting up fake ones
in the main test script, trying to run some tests, and then seeing if
those tests impacted the environment correctly. If they didn't, then we
write a message and manually call exit.
Instead, let's push these down into a sub-test, like many of the other
tests covering the framework itself. This has a few advantages:
- it does not pollute the test output with mention of skipped tests
(that we know are uninteresting -- the point of the test was to see
that these are skipped).
- when running in a TAP harness, we get a useful test failure message
(whereas when the script exits early, a tool like "prove" simply
says "Dubious, test returned 1").
- we do not have to worry about different test environments, such as
when GIT_TEST_FAIL_PREREQS_INTERNAL is set. Our sub-test helpers
already give us a known environment.
- the tests themselves are a bit easier to read, as we can just check
the test-framework output to see what happened (and get the usual
test_cmp diff if it failed)
A few notes on the implementation:
- we could do one sub-test per each individual test_expect_success. I
broke it up here into a few logical groups, as I think this makes it
more readable
- the original tests modified environment variables inside the test
bodies. Instead, I've used "true" as the body of a test we expect to
run and "false" otherwise. Technically this does not confirm that
the body of the "true" test actually ran. We are trusting the
framework output to believe that it truly ran, which is sufficient
for these tests. And I think the end result is much simpler to
follow.
- the nested_prereq test uses a few bare "test -f" calls; I converted
these to our usual test_path_is_* helpers while moving the code
around.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We check that test_when_finished cleans up after a test, and that it
runs even after a failure. Those two were originally adjacent, but got
split apart by the new test added in 477dcaddb6 (tests: do not let lazy
prereqs inside `test_expect_*` turn off tracing, 2020-03-26), and then
further by more lazy-prereq tests. Let's move them back together.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we are looking up an oid in an array, we obviously don't need to
write to the array. Let's mark it as const in the function interfaces,
as well as in the local variables we use to derference the void pointer
(note a few cases use pointers-to-pointers, so we mark everything
const).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
All of our callers are actually looking up an object_id, not a bare
hash. Likewise, the arrays they are looking in are actual arrays of
object_id (not just raw bytes of hashes, as we might find in a pack
.idx; those are handled by bsearch_hash()).
Using an object_id gives us more type safety, and makes the callers
slightly shorter. It also gets rid of the word "sha1" from several
access functions, though we could obviously also rename those with
s/sha1/hash/.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We store a struct for each directory we access under .git/rr-cache. The
structs are kept in an array sorted by the binary hash associated with
their name (and we do lookups with a binary search).
This works OK, but there are a few small downsides:
- the amount of code isn't huge, but it's more than we'd need using one
of our other stock data structures
- the insertion into a sorted array is quadratic (though in practice
it's unlikely anybody has enough conflicts for this to matter)
- it's intimately tied to the representation of an object hash. This
isn't a big deal, as the conflict ids we generate use the same hash,
but it produces a few awkward bits (e.g., we are the only user of
hash_pos() that is not using object_id).
Let's instead just treat the directory names as strings, and store them
in a strmap. This is less code, and removes the use of hash_pos().
Insertion is now non-quadratic, though we probably use a bit more
memory. Besides the hash table overhead, and storing hex bytes instead
of a binary hash, we actually store each name twice. Other code expects
to access the name of a rerere_dir struct from the struct itself, so we
need a copy there. But strmap keeps its own copy of the name, as well.
Using a bare hashmap instead of strmap means we could use the name for
both, but at the cost of extra code (e.g., our own comparison function).
Likewise, strmap has a feature to use a pointer to the in-struct name at
the cost of a little extra code. I didn't do either here, as simple code
seemed more important than squeezing out a few bytes of efficiency.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We check only that get_sha1_hex() doesn't complain, which means we'd
match an all-hex name with trailing cruft after it. This probably
doesn't matter much in practice, since there shouldn't be anything else
in the rr-cache directory, but it could possibly cause us to mix up sha1
and sha256 entries (which also shouldn't be intermingled, but could be
leftovers from a repository conversion).
Note that "get_sha1_hex()" is a confusing historical name. It is
actually using the_hash_algo, so it would be sha256 in a sha256 repo.
We'll switch to using parse_oid_hex(), because that conveniently
advances our pointer. But it also gets rid of the sha1 name. Arguably
it's a little funny to use "object_id" here for something that isn't
actually naming an object, but it's unlikely to be a problem (and is
contained in a single function).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In rerere_gc(), we walk over the .git/rr_cache directory and create a
struct for each entry we find. We feed any name we get from readdir() to
find_rerere_dir(), which then calls get_sha1_hex() on it (since we use
the binary hash as a lookup key). If that fails (i.e., the directory
name is not what we expected), it returns NULL. But the comment in
find_rerere_dir() says "BUG".
It _would_ be a bug for the call from new_rerere_id_hex(), the only
other code path, to fail here; it's generating the hex internally. But
the call in rerere_gc() is using it say "is this a plausible directory
name".
Let's instead have rerere_gc() do its own "is this plausible" check.
That has two benefits:
- we can now reliably BUG() inside find_rerere_dir(), which would
catch bugs in the other code path (and we now will never return NULL
from the function, which makes it easier to see that a rerere_id
struct will always have a non-NULL "collection" field).
- it makes the use of the binary hash an implementation detail of
find_rerere_dir(), not known by callers. That will free us up to
change it in a future patch.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
All of our callers have an object_id, and are just dereferencing the
hash field to pass to us. Let's take the actual object_id instead. We
still access the hash to pass to hash_pos, but it's a step in the right
direction.
This makes the callers slightly simpler, but also gets rid of the
untyped pointer, as well as the now-inaccurate name "sha1".
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix a bug in upload-pack.c that occurs when you combine partial
clone and uploadpack.packObjectsHook. You can reproduce it as
follows:
git clone -u 'git -c uploadpack.allowfilter '\
'-c uploadpack.packobjectshook=env '\
'upload-pack' --filter=blob:none --no-local \
src.git dst.git
Be careful with the line endings because this has a long quoted
string as the -u argument.
The error I get when I run this is:
Cloning into '/tmp/broken'...
remote: fatal: invalid filter-spec ''blob:none''
error: git upload-pack: git-pack-objects died with error.
fatal: git upload-pack: aborting due to possible repository corruption on the remote side.
remote: aborting due to possible repository corruption on the remote side.
fatal: early EOF
fatal: index-pack failed
The problem is caused by unneeded quoting.
This bug was already present in 10ac85c785 (upload-pack: add object
filtering for partial clone, 2017-12-08) when the server side filter
support was introduced. In fact, in 10ac85c785 this was broken
regardless of uploadpack.packObjectsHook. Then in 0b6069fe0a
(fetch-pack: test support excluding large blobs, 2017-12-08) the
quoting was removed but only behind a conditional that depends on
whether uploadpack.packObjectsHook is set.
Because uploadpack.packObjectsHook is apparently rarely used, nobody
noticed the problematic quoting could still happen.
Remove the conditional quoting and add a test for partial clone in
t5544-pack-objects-hook.
Signed-off-by: Jacob Vosmaer <jacob@gitlab.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We allow variadic macros in the code base, but only if there is fallback
code for platforms that lack it. This leads to some annoyances:
- the code is more complicated because of the fallbacks (e.g.,
trace_printf(), etc, is implemented twice with a set of parallel
wrappers).
- some constructs are just impossible and we've had to live without
them (e.g., a cross between FLEX_ALLOC and xstrfmt)
Since this feature is present in C99, we may be able to start counting
on it being available everywhere. Let's start with a weather balloon
patch to find out.
This patch makes the absolute minimal change by always setting
HAVE_VARIADIC_MACROS. If somebody runs into a platform where it's a
problem, they can undo it by commenting out the define. Likewise, if we
have to revert this, it would be quite unlikely to cause conflicts.
Once we feel comfortable that this is the right direction, then we can
start ripping out all the spots that actually look at the flag, and
removing the dead code.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The CI/PR GitHub Actions workflow uses the 'matrix' strategy for the
"windows-test", "vs-test", "regular" and "dockerized" jobs. The default
behaviour of GitHub Actions is to cancel all in-progress jobs in a
matrix if one of the job of the matrix fails [1].
This is not ideal as a failure early in a job, like during installation of
the build/test dependencies on a specific platform, leads to the
cancellation of all other jobs in the matrix.
Set the 'fail-fast' variable to 'false' for all four matrix jobs in the
workflow.
[1] https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#jobsjob_idstrategyfail-fast
Signed-off-by: Philippe Blain <levraiphilippeblain@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'./t1234-foo.sh --stress-jobs=X ...' is supposed to run that test
script in X parallel jobs, but the number of jobs specified on the
command line is entirely ignored if other '--stress'-related options
follow. I.e. both './t1234-foo.sh --stress-jobs=X --stress-limit=Y'
and './t1234-foo.sh --stress-jobs=X --stress' fall back to using twice
the number of CPUs parallel jobs instead.
The former has been broken since commit de69e6f6c9 (tests: let
--stress-limit=<N> imply --stress, 2019-03-03) [1], which started to
unconditionally overwrite the $stress variable holding the specified
number of jobs in its effort to imply '--stress'. The latter has been
broken since f545737144 (tests: introduce --stress-jobs=<N>,
2019-03-03), because it didn't consider that handling '--stress' will
overwrite that variable as well.
We could fix this by being more careful about (over)writing that
$stress variable and checking first whether it has already been set.
But I think it's cleaner to use a dedicated variable to hold the
number of specified parallel jobs, so let's do that instead.
[1] In de69e6f6c9 there was no '--stress-jobs=X' option yet, the
number of parallel jobs had to be specified via '--stress=X', so,
strictly speaking, de69e6f6c9 broke './t1234-foo.sh --stress=X
--stress-limit=Y'.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Remove the hidden "grep --debug" and "log --grep-debug" options added
in 17bf35a3c7 (grep: teach --debug option to dump the parse tree,
2012-09-13).
At the time these options seem to have been intended to go along with
a documentation discussion and to help the author of relevant tests to
perform ad-hoc debugging on them[1].
Reasons to want this gone:
1. They were never documented, and the only (rather trivial) use of
them in our own codebase for testing is something I removed back
in e01b4dab01 (grep: change non-ASCII -i test to stop using
--debug, 2017-05-20).
2. Googling around doesn't show any in-the-wild uses I could dig up,
and on the Git ML the only mentions after the original discussion
seem to have been when they came up in unrelated diff contexts, or
that test commit of mine.
3. An exception to that is c581e4a749 (grep: under --debug, show
whether PCRE JIT is enabled, 2019-08-18) where we added the
ability to dump out when PCREv2 has the JIT in effect.
The combination of that and my earlier b65abcafc7 (grep: use PCRE
v2 for optimized fixed-string search, 2019-07-01) means Git prints
this out in its most common in-the-wild configuration:
$ git log --grep-debug --grep=foo --grep=bar --grep=baz --all-match
pcre2_jit_on=1
pcre2_jit_on=1
pcre2_jit_on=1
[all-match]
(or
pattern_body<body>foo
(or
pattern_body<body>bar
pattern_body<body>baz
)
)
$ git grep --debug \( -e foo --and -e bar \) --or -e baz
pcre2_jit_on=1
pcre2_jit_on=1
pcre2_jit_on=1
(or
(and
patternfoo
patternbar
)
patternbaz
)
I.e. for each pattern we're considering for the and/or/--all-match
etc. debugging we'll now diligently spew out another identical line
saying whether the PCREv2 JIT is on or not.
I think that nobody's complained about that rather glaringly obviously
bad output says something about how much this is used, i.e. it's
not.
The need for this debugging aid for the composed grep/log patterns
seems to have passed, and the desire to dump the JIT config seems to
have been another one-off around the time we had JIT-related issues on
the PCREv2 codepath. That the original author of this debugging
facility seemingly hasn't noticed the bad output since then[2] is
probably some indicator.
1. https://lore.kernel.org/git/cover.1347615361.git.git@drmicha.warpmail.net/
2. https://lore.kernel.org/git/xmqqk1b8x0ac.fsf@gitster-ct.c.googlers.com/
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When an on-disk reverse index exists, there is no need to generate one
in memory. In fact, doing so can be slow, and require large amounts of
the heap.
Let's make sure that we treat the on-disk reverse index with precedence
(i.e., that when it exists, we don't bother trying to generate an
equivalent one in memory) by teaching Git how to conditionally die()
when generating a reverse index in memory.
Then, add a test to ensure that when (a) an on-disk reverse index
exists, and (b) when setting GIT_TEST_REV_INDEX_DIE_IN_MEMORY, that we
do not die, implying that we read from the on-disk one.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a new option that unconditionally enables the pack.writeReverseIndex
setting in order to run the whole test suite in a mode that generates
on-disk reverse indexes. Additionally, enable this mode in the second
run of tests under linux-gcc in 'ci/run-build-and-tests.sh'.
Once on-disk reverse indexes are proven out over several releases, we
can change the default value of that configuration to 'true', and drop
this patch.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the next patch, we'll add support for unconditionally enabling the
'pack.writeReverseIndex' setting with a new GIT_TEST_WRITE_REV_INDEX
environment variable.
This causes a little bit of fallout with tests that, for example,
compare the list of files in the pack directory being unprepared to see
.rev files in its output.
Those locations can be cleaned up to look for specific file extensions,
rather than take everything in the pack directory (for instance) and
then grep out unwanted items.
Once the pack.writeReverseIndex option has been thoroughly
tested, we will default it to 'true', removing GIT_TEST_WRITE_REV_INDEX,
and making it possible to revert this patch.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the pack.writeReverseIndex configuration is respected in both
'git index-pack' and 'git pack-objects' (and therefore, all of their
callers), we can safely advertise it for use in the git-config manual.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we have an implementation that can write the new reverse index
format, enable writing a .rev file in 'git pack-objects' by consulting
the pack.writeReverseIndex configuration variable.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach 'git index-pack' to optionally write and verify reverse index with
'--[no-]rev-index', as well as respecting the 'pack.writeReverseIndex'
configuration option.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To derive the filename for a .idx file, 'git index-pack' uses
derive_filename() to strip the '.pack' suffix and add the new suffix.
Prepare for stripping off suffixes other than '.pack' by making the
suffix to strip a parameter of derive_filename(). In order to make this
consistent with the "suffix" parameter which does not begin with a ".",
an additional check in derive_filename.
Suggested-by: Jeff King <peff@peff.net>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch prepares for callers to be able to write reverse index files
to disk.
It adds the necessary machinery to write a format-compliant .rev file
from within 'write_rev_file()', which is called from
'finish_tmp_packfile()'.
Similar to the process by which the reverse index is computed in memory,
these new paths also have to sort a list of objects by their offsets
within a packfile. These new paths use a qsort() (as opposed to a radix
sort), since our specialized radix sort requires a full revindex_entry
struct per object, which is more memory than we need to allocate.
The qsort is obviously slower, but the theoretical slowdown would
require a repository with a large amount of objects, likely implying
that the time spent in, say, pack-objects during a repack would dominate
the overall runtime.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Specify the format of the on-disk reverse index 'pack-*.rev' file, as
well as prepare the code for the existence of such files.
The reverse index maps from pack relative positions (i.e., an index into
the array of object which is sorted by their offsets within the
packfile) to their position within the 'pack-*.idx' file. Today, this is
done by building up a list of (off_t, uint32_t) tuples for each object
(the off_t corresponding to that object's offset, and the uint32_t
corresponding to its position in the index). To convert between pack and
index position quickly, this array of tuples is radix sorted based on
its offset.
This has two major drawbacks:
First, the in-memory cost scales linearly with the number of objects in
a pack. Each 'struct revindex_entry' is sizeof(off_t) +
sizeof(uint32_t) + padding bytes for a total of 16.
To observe this, force Git to load the reverse index by, for e.g.,
running 'git cat-file --batch-check="%(objectsize:disk)"'. When asking
for a single object in a fresh clone of the kernel, Git needs to
allocate 120+ MB of memory in order to hold the reverse index in memory.
Second, the cost to sort also scales with the size of the pack.
Luckily, this is a linear function since 'load_pack_revindex()' uses a
radix sort, but this cost still must be paid once per pack per process.
As an example, it takes ~60x longer to print the _size_ of an object as
it does to print that entire object's _contents_:
Benchmark #1: git.compile cat-file --batch <obj
Time (mean ± σ): 3.4 ms ± 0.1 ms [User: 3.3 ms, System: 2.1 ms]
Range (min … max): 3.2 ms … 3.7 ms 726 runs
Benchmark #2: git.compile cat-file --batch-check="%(objectsize:disk)" <obj
Time (mean ± σ): 210.3 ms ± 8.9 ms [User: 188.2 ms, System: 23.2 ms]
Range (min … max): 193.7 ms … 224.4 ms 13 runs
Instead, avoid computing and sorting the revindex once per process by
writing it to a file when the pack itself is generated.
The format is relatively straightforward. It contains an array of
uint32_t's, the length of which is equal to the number of objects in the
pack. The ith entry in this table contains the index position of the
ith object in the pack, where "ith object in the pack" is determined by
pack offset.
One thing that the on-disk format does _not_ contain is the full (up to)
eight-byte offset corresponding to each object. This is something that
the in-memory revindex contains (it stores an off_t in 'struct
revindex_entry' along with the same uint32_t that the on-disk format
has). Omit it in the on-disk format, since knowing the index position
for some object is sufficient to get a constant-time lookup in the
pack-*.idx file to ask for an object's offset within the pack.
This trades off between the on-disk size of the 'pack-*.rev' file for
runtime to chase down the offset for some object. Even though the lookup
is constant time, the constant is heavier, since it can potentially
involve two pointer walks in v2 indexes (one to access the 4-byte offset
table, and potentially a second to access the double wide offset table).
Consider trying to map an object's pack offset to a relative position
within that pack. In a cold-cache scenario, more page faults occur while
switching between binary searching through the reverse index and
searching through the *.idx file for an object's offset. Sure enough,
with a cold cache (writing '3' into '/proc/sys/vm/drop_caches' after
'sync'ing), printing out the entire object's contents is still
marginally faster than printing its size:
Benchmark #1: git.compile cat-file --batch-check="%(objectsize:disk)" <obj >/dev/null
Time (mean ± σ): 22.6 ms ± 0.5 ms [User: 2.4 ms, System: 7.9 ms]
Range (min … max): 21.4 ms … 23.5 ms 41 runs
Benchmark #2: git.compile cat-file --batch <obj >/dev/null
Time (mean ± σ): 17.2 ms ± 0.7 ms [User: 2.8 ms, System: 5.5 ms]
Range (min … max): 15.6 ms … 18.2 ms 45 runs
(Numbers taken in the kernel after cheating and using the next patch to
generate a reverse index). There are a couple of approaches to improve
cold cache performance not pursued here:
- We could include the object offsets in the reverse index format.
Predictably, this does result in fewer page faults, but it triples
the size of the file, while simultaneously duplicating a ton of data
already available in the .idx file. (This was the original way I
implemented the format, and it did show
`--batch-check='%(objectsize:disk)'` winning out against `--batch`.)
On the other hand, this increase in size also results in a large
block-cache footprint, which could potentially hurt other workloads.
- We could store the mapping from pack to index position in more
cache-friendly way, like constructing a binary search tree from the
table and writing the values in breadth-first order. This would
result in much better locality, but the price you pay is trading
O(1) lookup in 'pack_pos_to_index()' for an O(log n) one (since you
can no longer directly index the table).
So, neither of these approaches are taken here. (Thankfully, the format
is versioned, so we are free to pursue these in the future.) But, cold
cache performance likely isn't interesting outside of one-off cases like
asking for the size of an object directly. In real-world usage, Git is
often performing many operations in the revindex (i.e., asking about
many objects rather than a single one).
The trade-off is worth it, since we will avoid the vast majority of the
cost of generating the revindex that the extra pointer chase will look
like noise in the following patch's benchmarks.
This patch describes the format and prepares callers (like in
pack-revindex.c) to be able to read *.rev files once they exist. An
implementation of the writer will appear in the next patch, and callers
will gradually begin to start using the writer in the patches that
follow after that.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Abstract accesses to in-core revindex that allows enumerating
objects stored in a packfile in the order they appear in the pack,
in preparation for introducing an on-disk precomputed revindex.
* tb/pack-revindex-api: (21 commits)
for_each_object_in_pack(): clarify pack vs index ordering
pack-revindex.c: avoid direct revindex access in 'offset_to_pack_pos()'
pack-revindex: hide the definition of 'revindex_entry'
pack-revindex: remove unused 'find_revindex_position()'
pack-revindex: remove unused 'find_pack_revindex()'
builtin/gc.c: guess the size of the revindex
for_each_object_in_pack(): convert to new revindex API
unpack_entry(): convert to new revindex API
packed_object_info(): convert to new revindex API
retry_bad_packed_offset(): convert to new revindex API
get_delta_base_oid(): convert to new revindex API
rebuild_existing_bitmaps(): convert to new revindex API
try_partial_reuse(): convert to new revindex API
get_size_by_pos(): convert to new revindex API
show_objects_for_type(): convert to new revindex API
bitmap_position_packfile(): convert to new revindex API
check_object(): convert to new revindex API
write_reused_pack_verbatim(): convert to new revindex API
write_reused_pack_one(): convert to new revindex API
write_reuse_object(): convert to new revindex API
...
Update the Code-of-conduct to version 2.0 from the upstream (we've
been using version 1.4).
* ab/coc-update-to-2.0:
CoC: update to version 2.0 + local changes
CoC: explicitly take any whitespace breakage
CoC: Update word-wrapping to match upstream
Introduce two new ways to feed configuration variable-value pairs
via environment variables, and tweak the way GIT_CONFIG_PARAMETERS
encodes variable/value pairs to make it more robust.
* ps/config-env-pairs:
config: allow specifying config entries via envvar pairs
environment: make `getenv_safe()` a public function
config: store "git -c" variables using more robust format
config: parse more robust format in GIT_CONFIG_PARAMETERS
config: extract function to parse config pairs
quote: make sq_dequote_step() a public function
config: add new way to pass config via `--config-env`
git: add `--super-prefix` to usage string
A bit of code refactoring.
* cc/write-promisor-file:
pack-write: die on error in write_promisor_file()
fetch-pack: refactor writing promisor file
fetch-pack: rename helper to create_promisor_file()
"git bundle" learns "--stdin" option to read its refs from the
standard input. Also, it now does not lose refs whey they point
at the same object.
* jx/bundle:
bundle: arguments can be read from stdin
bundle: lost objects when removing duplicate pendings
test: add helper functions for git-bundle
Clean-up docs, codepaths and tests around mailmap.
* ab/mailmap: (22 commits)
shortlog: remove unused(?) "repo-abbrev" feature
mailmap doc + tests: document and test for case-insensitivity
mailmap tests: add tests for empty "<>" syntax
mailmap tests: add tests for whitespace syntax
mailmap tests: add a test for comment syntax
mailmap doc + tests: add better examples & test them
tests: refactor a few tests to use "test_commit --append"
test-lib functions: add an --append option to test_commit
test-lib functions: add --author support to test_commit
test-lib functions: document arguments to test_commit
test-lib functions: expand "test_commit" comment template
mailmap: test for silent exiting on missing file/blob
mailmap tests: get rid of overly complex blame fuzzing
mailmap tests: add a test for "not a blob" error
mailmap tests: remove redundant entry in test
mailmap tests: improve --stdin tests
mailmap tests: modernize syntax & test idioms
mailmap tests: use our preferred whitespace syntax
mailmap doc: start by mentioning the comment syntax
check-mailmap doc: note config options
...
"git fetch" learns to treat ref updates atomically in all-or-none
fashion, just like "git push" does, with the new "--atomic" option.
* ps/fetch-atomic:
fetch: implement support for atomic reference updates
fetch: allow passing a transaction to `s_update_ref()`
fetch: refactor `s_update_ref` to use common exit path
fetch: use strbuf to format FETCH_HEAD updates
fetch: extract writing to FETCH_HEAD
When more than one commit with the same patch ID appears on one
side, "git log --cherry-pick A...B" did not exclude them all when a
commit with the same patch ID appears on the other side. Now it
does.
* jk/log-cherry-pick-duplicate-patches:
patch-ids: handle duplicate hashmap entries
Prepare tests not to be affected by the name of the default branch
"git init" creates.
* js/default-branch-name-tests-final-stretch: (28 commits)
tests: drop prereq `PREPARE_FOR_MAIN_BRANCH` where no longer needed
t99*: adjust the references to the default branch name "main"
tests(git-p4): transition to the default branch name `main`
t9[5-7]*: adjust the references to the default branch name "main"
t9[0-4]*: adjust the references to the default branch name "main"
t8*: adjust the references to the default branch name "main"
t7[5-9]*: adjust the references to the default branch name "main"
t7[0-4]*: adjust the references to the default branch name "main"
t6[4-9]*: adjust the references to the default branch name "main"
t64*: preemptively adjust alignment to prepare for `master` -> `main`
t6[0-3]*: adjust the references to the default branch name "main"
t5[6-9]*: adjust the references to the default branch name "main"
t55[4-9]*: adjust the references to the default branch name "main"
t55[23]*: adjust the references to the default branch name "main"
t551*: adjust the references to the default branch name "main"
t550*: adjust the references to the default branch name "main"
t5503: prepare aligned comment for replacing `master` with `main`
t5[0-4]*: adjust the references to the default branch name "main"
t5323: prepare centered comment for `master` -> `main`
t4*: adjust the references to the default branch name "main"
...
After expiring a reflog and making a single commit, the reflog for
the branch would record a single entry that knows both @{0} and
@{1}, but we failed to answer "what commit were we on?", i.e. @{1}
* dl/reflog-with-single-entry:
refs: allow @{n} to work with n-sized reflog
refs: factor out set_read_ref_cutoffs()
"git diff" showed a submodule working tree with untracked cruft as
"Submodule commit <objectname>-dirty", but a natural expectation is
that the "-dirty" indicator would align with "git describe --dirty",
which does not consider having untracked files in the working tree
as source of dirtiness. The inconsistency has been fixed.
* sj/untracked-files-in-submodule-directory-is-not-dirty:
diff: do not show submodule with untracked files as "-dirty"
Warn loudly when the "pack-redundant" command, which has been left
stale with almost unusable performance issues, gets used, as we no
longer want to recommend its use (instead just "repack -d" instead).
* jc/deprecate-pack-redundant:
pack-redundant: gauge the usage before proposing its removal
Newline characters in the host and path part of git:// URL are
now forbidden.
* jk/forbid-lf-in-git-url:
fsck: reject .gitmodules git:// urls with newlines
git_connect_git(): forbid newlines in host and path
The implementation of "git branch --sort" wrt the detached HEAD
display has always been hacky, which has been cleaned up.
* ab/branch-sort:
branch: show "HEAD detached" first under reverse sort
branch: sort detached HEAD based on a flag
ref-filter: move ref_sorting flags to a bitfield
ref-filter: move "cmp_fn" assignment into "else if" arm
ref-filter: add braces to if/else if/else chain
branch tests: add to --sort tests
branch: change "--local" to "--list" in comment
File-level rename detection updates.
* en/diffcore-rename:
diffcore-rename: remove unnecessary duplicate entry checks
diffcore-rename: accelerate rename_dst setup
diffcore-rename: simplify and accelerate register_rename_src()
t4058: explore duplicate tree entry handling in a bit more detail
t4058: add more tests and documentation for duplicate tree entry handling
diffcore-rename: reduce jumpiness in progress counters
diffcore-rename: simplify limit check
diffcore-rename: avoid usage of global in too_many_rename_candidates()
diffcore-rename: rename num_create to num_destinations
Rename detection is added to the "ORT" merge strategy.
* en/merge-ort-3:
merge-ort: add implementation of type-changed rename handling
merge-ort: add implementation of normal rename handling
merge-ort: add implementation of rename collisions
merge-ort: add implementation of rename/delete conflicts
merge-ort: add implementation of both sides renaming differently
merge-ort: add implementation of both sides renaming identically
merge-ort: add basic outline for process_renames()
merge-ort: implement compare_pairs() and collect_renames()
merge-ort: implement detect_regular_renames()
merge-ort: add initial outline for basic rename detection
merge-ort: add basic data structures for handling renames
"git mktag" validates its input using its own rules before writing
a tag object---it has been updated to share the logic with "git
fsck".
* ab/mktag: (23 commits)
mktag: add a --[no-]strict option
mktag: mark strings for translation
mktag: convert to parse-options
mktag: allow omitting the header/body \n separator
mktag: allow turning off fsck.extraHeaderEntry
fsck: make fsck_config() re-usable
mktag: use fsck instead of custom verify_tag()
mktag: use puts(str) instead of printf("%s\n", str)
mktag: remove redundant braces in one-line body "if"
mktag: use default strbuf_read() hint
mktag tests: test verify_object() with replaced objects
mktag tests: improve verify_object() test coverage
mktag tests: test "hash-object" compatibility
mktag tests: stress test whitespace handling
mktag tests: run "fsck" after creating "mytag"
mktag tests: don't create "mytag" twice
mktag tests: don't redirect stderr to a file needlessly
mktag tests: remove needless SHA-1 hardcoding
mktag tests: use "test_commit" helper
mktag tests: don't needlessly use a subshell
...
Improve the support for invalid UTF-8 haystacks given a non-ASCII
needle when using the PCREv2 backend.
This is a more complete fix for a bug I started to fix in
870eea8166 (grep: do not enter PCRE2_UTF mode on fixed matching,
2019-07-26), now that PCREv2 has the PCRE2_MATCH_INVALID_UTF mode we
can make use of it.
This fixes the sort of case described in 8a5999838e (grep: stess test
PCRE v2 on invalid UTF-8 data, 2019-07-26), i.e.:
- The subject string is non-ASCII (e.g. "ævar")
- We're under a is_utf8_locale(), e.g. "en_US.UTF-8", not "C"
- We are using --ignore-case, or we're a non-fixed pattern
If those conditions were satisfied and we matched found non-valid
UTF-8 data PCREv2 might bark on it, in practice this only happened
under the JIT backend (turned on by default on most platforms).
Ultimately this fixes a "regression" in b65abcafc7 ("grep: use PCRE v2
for optimized fixed-string search", 2019-07-01), I'm putting that in
scare-quotes because before then we wouldn't properly support these
complex case-folding, locale etc. cases either, it just broke in
different ways.
There was a bug related to this the PCRE2_NO_START_OPTIMIZE flag fixed
in PCREv2 10.36. It can be worked around by setting the
PCRE2_NO_START_OPTIMIZE flag. Let's do that in those cases, and add
tests for the bug.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As noted in [1] when I originally added this test in [2] the test was
completely broken as it lacked a redirect[3]. I now think this whole
thing is overly fragile. Let's only test if we have a segfault here.
Before this the first test's "test_cmp" was pretty meaningless. We
were only testing if PCREv2 was so broken that it would spew out
something completely unrelated on stdout, which isn't very plausible.
In the second test we're relying on PCREv2 forever holding to the
current behavior of the PCRE_UTF8 flag, as opposed to learning some
optimistic graceful fallback to PCRE2_MATCH_INVALID_UTF in the
future. If that happens having this test broken under bisecting would
suck.
A follow-up commit will actually test this case in a meaningful way
under the PCRE2_MATCH_INVALID_UTF flag. Let's run this one
unconditionally, and just make sure we don't segfault.
1. e714b898c6 (t7812: expect failure for grep -i with invalid UTF-8
data, 2019-11-29)
2. 8a5999838e (grep: stess test PCRE v2 on invalid UTF-8 data,
2019-07-26)
3. c74b3cbb83 (t7812: add missing redirects, 2019-11-26)
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
