Commit Graph

56610 Commits

Author SHA1 Message Date
Quentin Nerden
bfc8c84ed5 docs: git-clone: list short form of options first
List the short form of options (e.g.: '-l') before the long form (e.g.
'--local').
This is to match the doc of git-add, git-commit, git-clean, git-branch...

Signed-off-by: Quentin Nerden <quentin.nerden@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:11:40 -07:00
Quentin Nerden
3711d1cd89 docs: git-clone: refer to long form of options
To make the doc of git-clone easier to read,
refer to the long form of the options
(it is easier to guess what '--verbose' is doing than '-v').

Signed-off-by: Quentin Nerden <quentin.nerden@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:11:38 -07:00
Rohit Ashiwal
dcb500dc16 cherry-pick/revert: advise using --skip
The previous commit introduced a --skip flag for cherry-pick and
revert. Update the advice messages, to tell users about this less
cumbersome way of skipping commits. Also add tests to ensure
everything is working fine.

Signed-off-by: Rohit Ashiwal <rohit.ashiwal265@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:08:08 -07:00
Rohit Ashiwal
de81ca3f36 cherry-pick/revert: add --skip option
git am or rebase have a --skip flag to skip the current commit if the
user wishes to do so. During a cherry-pick or revert a user could
likewise skip a commit, but needs to use 'git reset' (or in the case
of conflicts 'git reset --merge'), followed by 'git (cherry-pick |
revert) --continue' to skip the commit. This is more annoying and
sometimes confusing on the users' part. Add a `--skip` option to make
skipping commits easier for the user and to make the commands more
consistent.

In the next commit, we will change the advice messages hence finishing
the process of teaching revert and cherry-pick "how to skip commits".

Signed-off-by: Rohit Ashiwal <rohit.ashiwal265@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:08:08 -07:00
Rohit Ashiwal
265ab48f26 sequencer: use argv_array in reset_merge
Avoid using magic numbers for array size and index under `reset_merge`
function. Use `argv_array` instead. This will make code shorter and
easier to extend.

Signed-off-by: Rohit Ashiwal <rohit.ashiwal265@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:08:08 -07:00
Rohit Ashiwal
918d1e6ed8 sequencer: rename reset_for_rollback to reset_merge
We are on a path to teach cherry-pick/revert how to skip commits. To
achieve this, we could really make use of existing functions.
reset_for_rollback is one such function, but the name does not
intuitively suggest to use it to reset a merge, which it was born to
perform, see 539047c ("revert: introduce --abort to cancel a failed
cherry-pick", 2011-11-23). Change the name to reset_merge to make
it more intuitive.

Signed-off-by: Rohit Ashiwal <rohit.ashiwal265@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:08:07 -07:00
Rohit Ashiwal
6a1f9046a4 sequencer: add advice for revert
In the case of merge conflicts, while performing a revert, we are
currently advised to use `git cherry-pick --<sequencer-options>`.
Introduce a separate advice message for `git revert`. Also change
the signature of `create_seq_dir` to handle which advice to display
selectively.

Signed-off-by: Rohit Ashiwal <rohit.ashiwal265@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 12:08:07 -07:00
Jeff King
ac093d5508 t5703: use test_commit_bulk
There are two loops that create 33 commits each using test_commit. Using
test_commit_bulk speeds this up from:

  Benchmark #1: ./t5703-upload-pack-ref-in-want.sh --root=/var/ram/git-tests
    Time (mean ± σ):      2.142 s ±  0.161 s    [User: 1.136 s, System: 0.974 s]
    Range (min … max):    1.903 s …  2.401 s    10 runs

to:

  Benchmark #1: ./t5703-upload-pack-ref-in-want.sh --root=/var/ram/git-tests
    Time (mean ± σ):      1.440 s ±  0.114 s    [User: 737.7 ms, System: 615.4 ms]
    Range (min … max):    1.230 s …  1.604 s    10 runs

for an average savings of almost 33%.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 10:11:54 -07:00
Jeff King
9516345ea5 t5702: use test_commit_bulk
There are two loops that create 32 commits each using test_commit. Using
test_commit_bulk speeds this up from:

  Benchmark #1: ./t5702-protocol-v2.sh --root=/var/ram/git-tests
    Time (mean ± σ):      5.409 s ±  0.513 s    [User: 2.382 s, System: 2.466 s]
    Range (min … max):    4.633 s …  5.927 s    10 runs

to:

  Benchmark #1: ./t5702-protocol-v2.sh --root=/var/ram/git-tests
    Time (mean ± σ):      3.956 s ±  0.242 s    [User: 1.775 s, System: 1.627 s]
    Range (min … max):    3.449 s …  4.239 s    10 runs

for an average savings of over 25%.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 10:11:54 -07:00
Jeff King
737b19b50c t3311: use test_commit_bulk
One of the tests in t3311 creates 300 commits by running "test_commit"
in a loop. This requires 900 processes. Instead, we can use
test_commit_bulk to do it with only four. This improves the runtime of
the script from:

  Benchmark #1: ./t3311-notes-merge-fanout.sh --root=/var/ram/git-tests
    Time (mean ± σ):      5.821 s ±  0.691 s    [User: 3.146 s, System: 2.782 s]
    Range (min … max):    4.783 s …  6.841 s    10 runs

to:

  Benchmark #1: ./t3311-notes-merge-fanout.sh --root=/var/ram/git-tests
    Time (mean ± σ):      1.743 s ±  0.116 s    [User: 1.144 s, System: 0.691 s]
    Range (min … max):    1.629 s …  1.994 s    10 runs

for an average speedup of over 70%.

Unfortunately we still have to run 300 instances of "git notes add",
since the point is to test the fanout that comes from adding notes one
by one.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 10:11:54 -07:00
Jeff King
1ac96cd1ba t5310: increase the number of bitmapped commits
The bitmap index we compute in t5310 has only 20 commits in it. This
gives poor coverage of bitmap_writer_select_commits(), which simply
writes a bitmap for everything when there are fewer than 100 commits.

Let's bump the number of commits in the test to cover the more complex
code paths (this does drop coverage of the individual lines of the
trivial path, but the complex path does everything it does and more).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 10:11:54 -07:00
Jeff King
b1c36cb849 test-lib: introduce test_commit_bulk
Some tests need to create a string of commits. Doing this with
test_commit is very heavy-weight, as it needs at least one process per
commit (and in fact, uses several).

For bulk creation, we can do much better by using fast-import, but it's
often a pain to generate the input. Let's provide a helper to do so.

We'll use t5310 as a guinea pig, as it has three 10-commit loops. Here
are hyperfine results before and after:

  [before]
  Benchmark #1: ./t5310-pack-bitmaps.sh --root=/var/ram/git-tests
    Time (mean ± σ):      2.846 s ±  0.305 s    [User: 3.042 s, System: 0.919 s]
    Range (min … max):    2.250 s …  3.210 s    10 runs

  [after]
  Benchmark #1: ./t5310-pack-bitmaps.sh --root=/var/ram/git-tests
    Time (mean ± σ):      2.210 s ±  0.174 s    [User: 2.570 s, System: 0.604 s]
    Range (min … max):    1.999 s …  2.590 s    10 runs

So we're over 20% faster, while making the callers slightly shorter. We
added a lot more lines in test-lib-function.sh, of course, and the
helper is way more featureful than we need here. But my hope is that it
will be flexible enough to use in more places.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-02 10:08:40 -07:00
Derrick Stolee
3612c2334a t5319: use 'test-tool path-utils' instead of 'ls -l'
Using 'ls -l' and parsing the columns to find file sizes is
problematic when the platform could report the owner as a name
with spaces. Instead, use the 'test-tool path-utils file-size'
command to list only the sizes.

Reported-by: Johannes Sixt <j6t@kdbg.org>
Helped-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:47:15 -07:00
brian m. carlson
f6c9b474a9 t2203: avoid hard-coded object ID values
In order to make this test work with multiple hash algorithms, compute
the object ID used in this test instead of hard-coding it.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:19 -07:00
brian m. carlson
bfefd5202b t1710: make hash independent
This test uses several index hashes, which necessarily depend on the
version of the index and the hash algorithm in use.  Use test_oid_cache
to provide values for these for both SHA-1 and SHA-256.  Also, compute
an object ID and use $EMPTY_BLOB to make the remainder of the tests
independent of the hash algorithm in use.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:19 -07:00
brian m. carlson
2e306f6c4f t1007: remove SHA1 prerequisites
Update this test to use test_oid_cache to specify the object IDs for
both SHA-1 and SHA-256.  Since this test now works with both algorithms,
remove the SHA1 prerequisite.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:19 -07:00
brian m. carlson
b0d3c42eb4 t0090: make test pass with SHA-256
One assertion of this test checks for a shrinking cache tree.  The
initial index contains a cache tree with two directory names but no
object ID, and the second index contains a cache tree with an object ID
but no directory name.

With SHA-1, the second index is smaller than the first, because the
directory information stored takes more than the 20 bytes of an SHA-1
hash, but with SHA-256, the hash is longer, and the test fails the
assertion that the second index is smaller than the first.

To address this issue, increase the length of the subdirectory name to
ensure that the cache tree does indeed shrink in size regardless of the
algorithm in use.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:19 -07:00
brian m. carlson
4772b33878 t0027: make hash size independent
Several parts of this test generate files that have specific hard-coded
object IDs in them.  We don't really care about what the object ID in
question is, so we turn them all to zeros.

However, because some of these values are fixed and some are generated,
they can be of different lengths, which causes problems when running
with SHA-256.  Furthermore, some assertions in this test use only fixed
object IDs and some use both fixed and generated ones, so converting
only the expected results fixes some tests while breaking others.
Convert both actual and expected object IDs to the all-zeros object ID
of the appropriate length to ensure that the test passes when using
SHA-256.

The astute observer will notice that both tr and sed are used here.
Converting the tr call to a sed y/// command looks logical at first, but
it isn't possible because POSIX doesn't allow escapes in y/// commands
other than "\\" and "\n".

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
brian m. carlson
0795aed0a0 t6030: make test work with SHA-256
Compute several object ID values instead of hard-coding them, and use
test_oid_to_path to cleanly produce a path for an object.

Note that the bisect code which is tested here remains sensitive to the
hash algorithm in use because it uses the object ID to disambiguate
between two equidistant commits.  Fortunately, SHA-1 and SHA-256
disambiguate identically in the cases we care about, so there is no need
to modify the test to accommodate this situation.  However, if a further
hash algorithm change occurs, this test may require some restructuring.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
brian m. carlson
0b78a1b22a t5000: make hash independent
This test uses a stub of a very large (64 GB) object to test our
generation of tar archives.  In doing so, it uses the object ID of the
object so it can insert it into the database properly.  Look up these
values using test_oid.  Restructure the test slightly to use
test_oid_in_path.

Since we care about the object, not how it is named in a particular hash
algorithm, rename it to "huge-object", which is shorter and more
descriptive.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
brian m. carlson
8de6b383c1 t1450: make hash size independent
Replace several hard-coded full and partial object IDs with variables or
computed values.  Create junk data to stuff inside an invalid tree that
can be either 20 or 32 bytes long.  Compute a binary all-zeros object ID
instead of hard-coding a 20-byte length.

Additionally, compute various object IDs by using test_oid and
$EMPTY_BLOB so that this test works with multiple hash algorithms.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
brian m. carlson
ff4cb42e77 t1410: make hash size independent
Instead of parsing object IDs using fixed-length shell patterns, use cut
to extract the first two characters of an object ID in addition to the
test helper for object paths.  Update another test to look up an
appropriate object ID fragment from the all-zeros object ID instead of
hardcoding the value.

Although the test for parsing reflogs at BUFSIZ boundaries passes, mark
it with the SHA1 prerequisite, as it doesn't currently usefully test
anything when using a hash longer than 20 bytes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
brian m. carlson
56d8892459 t: add helper to convert object IDs to paths
There are several places in our testsuite where we want to insert a
slash after an object ID to make it into a path we can reference under
.git/objects, and we have various ways of doing so.  Add a helper to
provide a standard way of doing this that works for all size hashes.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 13:28:18 -07:00
Phillip Wood
e981bf7525 git-prompt: improve cherry-pick/revert detection
If the user commits or resets a conflict resolution in the middle of a
sequence of cherry-picks or reverts then CHERRY_PICK_HEAD/REVERT_HEAD
will be removed and so in the absence of those files we need to check
.git/sequencer/todo to see if there is a cherry-pick or revert in
progress.

Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 12:39:44 -07:00
Michael Platings
78fafbb280 t8014: remove unnecessary braces
Signed-off-by: Michael Platings <michael@platin.gs>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 11:05:51 -07:00
SZEDER Gábor
1a64e07d23 Document that 'git -C ""' works and doesn't change directory
It's been behaving so since 6a536e2076 (git: treat "git -C '<path>'"
as a no-op when <path> is empty, 2015-03-06).

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 10:42:49 -07:00
Eric Wong
7328482253 repack: disable bitmaps-by-default if .keep files exist
Bitmaps aren't useful with multiple packs, and users with
.keep files ended up with redundant packs when bitmaps
got enabled by default in bare repos.

So detect when .keep files exist and stop enabling bitmaps
by default in that case.

Wasteful (but otherwise harmless) race conditions with .keep files
documented by Jeff King still apply and there's a chance we'd
still end up with redundant data on the FS:

  https://public-inbox.org/git/20190623224244.GB1100@sigill.intra.peff.net/

v2: avoid subshell in test case, be multi-index aware

Fixes: 36eba0323d ("repack: enable bitmaps by default on bare repos")
Signed-off-by: Eric Wong <e@80x24.org>
Helped-by: Jeff King <peff@peff.net>
Reported-by: Janos Farkas <chexum@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 10:29:08 -07:00
Christian Couder
fbec05c210 t0016: add 'remove' subcommand test
Testing the 'remove' subcommand was forgotten when t0016
was created. Let's fix that.

Helped-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 10:26:33 -07:00
Christian Couder
84f559f750 test-oidmap: remove 'add' subcommand
The 'add' subcommand is useless as it is mostly identical
to the 'put' subcommand, so let's remove it.

Helped-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 10:26:28 -07:00
Jeff King
39b44ba771 check_everything_connected: assume alternate ref tips are valid
When we receive a remote ref update to sha1 "X", we want to check that
we have all of the objects needed by "X". We can assume that our
repository is not currently corrupted, and therefore if we have a ref
pointing at "Y", we have all of its objects. So we can stop our
traversal from "X" as soon as we hit "Y".

If we make the same non-corruption assumption about any repositories we
use to store alternates, then we can also use their ref tips to shorten
the traversal.

This is especially useful when cloning with "--reference", as we
otherwise do not have any local refs to check against, and have to
traverse the whole history, even though the other side may have sent us
few or no objects. Here are results for the included perf test (which
shows off more or less the maximal savings, getting one new commit and
sharing the whole history):

Test                        HEAD^             HEAD
--------------------------------------------------------------------
[on git.git]
5600.3: clone --reference   2.94(2.86+0.08)   0.09(0.08+0.01) -96.9%
[on linux.git]
5600.3: clone --reference   45.74(45.34+0.41)   0.36(0.30+0.08) -99.2%

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 10:11:09 -07:00
Jeff King
709dfa6990 object-store.h: move for_each_alternate_ref() from transport.h
There's nothing inherently transport-related about enumerating the
alternate ref tips. The code has lived in transport.[ch] because the
only use so far had been advertising available tips during transport.
But it could be used for more, and a future patch will teach rev-list to
access these refs.

Let's move it alongside the other alt-odb code, declaring it in
object-store.h with the implementation in sha1-file.c.

This lets us drop the inclusion of transport.h from receive-pack, which
perhaps shows how it was misplaced (though receive-pack is about
transporting objects, transport.h is mostly about the client side).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 09:47:29 -07:00
Johannes Schindelin
906b63942a rebase --am: ignore rebase.rescheduleFailedExec
The `exec` command is specific to the interactive backend, therefore it
does not make sense for non-interactive rebases to heed that config
setting.

We still want to error out if a non-interactive rebase is started with
`--reschedule-failed-exec`, of course.

Reported by Vas Sudanagunta via:
https://github.com/git/git/commit/969de3ff0e0#commitcomment-33257187

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-07-01 09:43:49 -07:00
Carmine Zaccagnino
820419f7c9
l10n: it.po: remove an extra space
Remove an extra space between the dashes and the start of the
"abort" option.

Signed-off-by: Carmine Zaccagnino <carmine@carminezacc.com>
Signed-off-by: Alessandro Menti <alessandro.menti@alessandromenti.it>
2019-06-29 07:51:37 +02:00
Jeff King
07a54dc307 blame: drop some unused function parameters
These unused parameters were introduced recently as part of the
br/blame-ignore topic. I assume they are not indicative of bugs, but are
just leftovers from the development process (they were introduced by the
series but not used in any of its iterations).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-28 10:13:56 -07:00
Nguyễn Thái Ngọc Duy
663d25018f t7814: do not generate same commits in different repos
t7814 has repo tree like this

  initial-repo
    submodule
      sub

In each repo 'submodule' and 'sub', a commit is made to add the same
initial file 'a' with the same message 'add a'. If tests run fast
enough, the two commits are made in the same second, resulting
identical commits.

There is nothing wrong with that per-se. But it could make the test
flaky. Currently all submodule odbs are merged back in the main
one (because we can't, or couldn't, access separate submodule repos
otherwise). But eventually we need to access objects from the right
repo.

Because the same commit could sometimes be present in both 'submodule'
and 'sub', if there is a bug looking up objects in the wrong repo,
sometimes it will go unnoticed because it finds the needed object in the
wrong repo anyway.

Fix this by changing commit time after every commit. This makes all
commits unique. Of course there are still identical blobs in different
repos, but because we often lookup commit first, then tree and blob,
unique commits are already quite safe.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-28 09:32:23 -07:00
Taylor Blau
b31e2680c4 ref-filter.c: find disjoint pattern prefixes
Since cfe004a5a9 (ref-filter: limit traversal to prefix, 2017-05-22),
the ref-filter code has sought to limit the traversals to a prefix of
the given patterns.

That code stopped short of handling more than one pattern, because it
means invoking 'for_each_ref_in' multiple times. If we're not careful
about which patterns overlap, we will output the same refs multiple
times.

For instance, consider the set of patterns 'refs/heads/a/*',
'refs/heads/a/b/c', and 'refs/tags/v1.0.0'. If we naïvely ran:

  for_each_ref_in("refs/heads/a/*", ...);
  for_each_ref_in("refs/heads/a/b/c", ...);
  for_each_ref_in("refs/tags/v1.0.0", ...);

we would see 'refs/heads/a/b/c' (and everything underneath it) twice.

Instead, we want to partition the patterns into disjoint sets, where we
know that no ref will be matched by any two patterns in different sets.
In the above, these are:

  - {'refs/heads/a/*', 'refs/heads/a/b/c'}, and
  - {'refs/tags/v1.0.0'}

Given one of these disjoint sets, what is a suitable pattern to pass to
'for_each_ref_in'? One approach is to compute the longest common prefix
over all elements in that disjoint set, and let the caller cull out the
refs they didn't want. Computing the longest prefix means that in most
cases, we won't match too many things the caller would like to ignore.

The longest common prefixes of the above are:

  - {'refs/heads/a/*', 'refs/heads/a/b/c'} -> refs/heads/a/*
  - {'refs/tags/v1.0.0'}                   -> refs/tags/v1.0.0

We instead invoke:

  for_each_ref_in("refs/heads/a/*", ...);
  for_each_ref_in("refs/tags/v1.0.0", ...);

Which provides us with the refs we were looking for with a minimal
amount of extra cruft, but never a duplicate of the ref we asked for.

Implemented here is an algorithm which accomplishes the above, which
works as follows:

  1. Lexicographically sort the given list of patterns.

  2. Initialize 'prefix' to the empty string, where our goal is to
     build each element in the above set of longest common prefixes.

  3. Consider each pattern in the given set, and emit 'prefix' if it
     reaches the end of a pattern, or touches a wildcard character. The
     end of a string is treated as if it precedes a wildcard. (Note that
     there is some room for future work to detect that, e.g., 'a?b' and
     'abc' are disjoint).

  4. Otherwise, recurse on step (3) with the slice of the list
     corresponding to our current prefix (i.e., the subset of patterns
     that have our prefix as a literal string prefix.)

This algorithm is 'O(kn + n log(n))', where 'k' is max(len(pattern)) for
each pattern in the list, and 'n' is len(patterns).

By discovering this set of interesting patterns, we reduce the runtime
of multi-pattern 'git for-each-ref' (and other ref traversals) from
O(N) to O(n log(N)), where 'N' is the total number of packed references.

Running 'git for-each-ref refs/tags/a refs/tags/b' on a repository with
10,000,000 refs in 'refs/tags/huge-N', my best-of-five times drop from:

  real    0m5.805s
  user    0m5.188s
  sys     0m0.468s

to:

  real    0m0.001s
  user    0m0.000s
  sys     0m0.000s

On linux.git, the times to dig out two of the latest -rc tags drops from
0.002s to 0.001s, so the change on repositories with fewer tags is much
less noticeable.

Co-authored-by: Jeff King <peff@peff.net>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 13:14:06 -07:00
SZEDER Gábor
5b12e3123b progress: use term_clear_line()
To make sure that the previously displayed progress line is completely
covered up when the new line is shorter, commit 545dc345eb (progress:
break too long progress bar lines, 2019-04-12) added a bunch of
calculations to figure out how many characters it needs to overwrite
with spaces.

Use the just introduced term_clear_line() helper function to, well,
clear the last line, making all these calculations unnecessary, and
thus simplifying the code considerably.

Three tests in 't5541-http-push-smart.sh' 'grep' for specific text
shown in the progress lines at the beginning of the line, but now
those lines begin either with the ANSI escape sequence or with the
terminal width worth of space characters clearing the line.  Relax the
'grep' patterns to match anywhere on the line.  Note that only two of
these three tests fail without relaxing their 'grep' pattern, but the
third looks for the absence of the pattern, so it still succeeds, but
without the adjustment would potentially hide future regressions.

Note also that with this change we no longer need the length of the
previously displayed progress line, so the strbuf added to 'struct
progress' in d53ba841d4 (progress: assemble percentage and counters in
a strbuf before printing, 2019-04-05) is not strictly necessary
anymore.  We still keep it, though, as it avoids allocating and
releasing a strbuf each time the progress is updated.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:58:41 -07:00
SZEDER Gábor
d7d90885e0 rebase: fix garbled progress display with '-x'
When running a command with the 'exec' instruction during an
interactive rebase session, or for a range of commits using 'git
rebase -x', the output can be a bit garbled when the name of the
command is short enough:

  $ git rebase -x true HEAD~5
  Executing: true
  Executing: true
  Executing: true
  Executing: true
  Executing: true)
  Successfully rebased and updated refs/heads/master.

Note the ')' at the end of the last line.  It gets more garbled as the
range of commits increases:

  $ git rebase -x true HEAD~50
  Executing: true)
  [ repeated 3 more times ]
  Executing: true0)
  [ repeated 44 more times ]
  Executing: true00)
  Successfully rebased and updated refs/heads/master.

Those extra numbers and ')' are remnants of the previously displayed
"Rebasing (N/M)" progress lines that are usually completely
overwritten by the "Executing: <cmd>" lines, unless 'cmd' is short and
the "N/M" part is long.

Make sure that the previously displayed "Rebasing (N/M)" line is
cleared by using the term_clear_line() helper function added in the
previous patch.  Do so only when not being '--verbose', because in
that case these "Rebasing (N/M)" lines are not printed as progress
(i.e. as lines with '\r' at the end), but as "regular" output (with
'\n' at the end).

A couple of other rebase commands print similar messages, e.g.
"Stopped at <abbrev-oid>... <subject>" for the 'edit' or 'break'
commands, or the "Successfully rebased and updated <full-ref>." at the
very end.  These are so long that they practically always overwrite
that "Rebasing (N/M)" progress line, but let's be prudent, and clear
the last line before printing these, too.

In 't3420-rebase-autostash.sh' two helper functions prepare the
expected output of four tests that check the full output of 'git
rebase' and thus are affected by this change, so adjust their
expectations to account for the new line clearing.

Note that this patch doesn't completely eliminate the possibility of
similar garbled outputs, e.g. some error messages from rebase or the
"Auto-merging <file>" message from within the depths of the merge
machinery might not be long enough to completely cover the last
"Rebasing (N/M)" line.  This patch doesn't do anything about them,
because dealing with them individually would result in way too much
churn, while having a catch-all term_clear_line() call in the common
code path of pick_commits() would hide the "Rebasing (N/M)" line way
too soon, and it would either flicker or be invisible.

Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:58:20 -07:00
Johannes Schindelin
94238859b9 mingw: use Unicode functions explicitly
Many Win32 API functions actually exist in two variants: one with
the `A` suffix that takes ANSI parameters (`char *` or `const char *`)
and one with the `W` suffix that takes Unicode parameters (`wchar_t *`
or `const wchar_t *`).

The ANSI variant assumes that the strings are encoded according to
whatever is the current locale. This is not what Git wants to use on
Windows: we assume that `char *` variables point to strings encoded in
UTF-8.

There is a pseudo UTF-8 locale on Windows, but it does not work
as one might expect. In addition, if we overrode the user's locale, that
would modify the behavior of programs spawned by Git (such as editors,
difftools, etc), therefore we cannot use that pseudo locale.

Further, it is actually highly encouraged to use the Unicode versions
instead of the ANSI versions, so let's do precisely that.

Note: when calling the Win32 API functions _without_ any suffix, it
depends whether the `UNICODE` constant is defined before the relevant
headers are #include'd. Without that constant, the ANSI variants are
used. Let's be explicit and avoid that ambiguity.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:56:15 -07:00
Johannes Schindelin
39a98e9b68 mingw: get pw_name in UTF-8 format
Previously, we would have obtained the user name encoded in whatever the
current code page is.

Note: the "user name" here does not denote the full name but instead the
short logon name.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:56:13 -07:00
Cesar Eduardo Barros
fe90397604 mingw: embed a manifest to trick UAC into Doing The Right Thing
On Windows >= Vista, not having an application manifest with a
requestedExecutionLevel can cause several kinds of confusing behavior.

The first and more obvious behavior is "Installer Detection" of the
"User Account Control" (also known as "UAC") feature, where Windows
sometimes decides (by looking at things like the file name and even
sequences of bytes within the executable) that an executable is an
installer and should run elevated (causing the well-known popup dialog
to appear). In Git's context, subcommands such as "git patch-id" or "git
update-index" fall prey to this behavior.

The second and more confusing behavior is "File Virtualization". It
means that when files are written without having write permission, it
does not fail (as expected), but they are instead redirected to
somewhere else. When the files are read, the original contents are
returned, though, not the ones that were just written somewhere else.
Even more confusing, not all write accesses are redirected; Trying to
write to write-protected .exe files, for example, will fail instead of
redirecting.

In addition to being unwanted behavior, File Virtualization causes
dramatic slowdowns in Git (see for instance
http://code.google.com/p/msysgit/issues/detail?id=320).

A third unwanted behavior of Windows >= Vista is that it lies about the
Windows version when calling `GetWindowsVersionEx()`.

There are two ways to prevent these unwanted behaviors: Either you embed
an application manifest (which really is an XML document conforming to a
specific schema) within all your executables, or you add an external
manifest (a file with the same name followed by `.manifest`) to all your
executables. Since Git's builtins are hardlinked (or copied), it is
simpler and more robust to embed a manifest.

Recent enough MSVC compilers already embed a working internal manifest,
and building with mingw-w64 (which is the case in Git for Windows' SDK)
does it, too, but for MinGW you have to do so by hand.

In any case, it is better to be explicit about this manifest, that way
changes in the compiler toolchain won't surprise us (as mingw-w64 once
did when it broke `GetWindowsVersionEx()` by mistake).

References:
  - New UAC Technologies for Windows Vista
    http://msdn.microsoft.com/en-us/library/bb756960.aspx
  - Create and Embed an Application Manifest (UAC)
    http://msdn.microsoft.com/en-us/library/bb756929.aspx

Signed-off-by: Cesar Eduardo Barros <cesarb@cesarb.net>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:55:45 -07:00
Johannes Schindelin
a1c5e906c8 mingw: enable stack smashing protector
To reduce Git for Windows' attack surface, we started using the Address
Space Layout Randomization and Data Execution Prevention features in
ce6a158561 (mingw: enable DEP and ASLR, 2019-05-08).

To remove yet another attack vector, let's make use of gcc's stack
smashing protector that helps detect stack buffer overruns early.

Rather than using -fstack-protector, we use -fstack-protector-strong
because on Windows: The latter appears to strike a better balance
between the performance impact and the provided safety.

In a non-scientific test (time git log --grep=is -p), best of 5 timings
went from 23.009s to 22.997s, i.e. the performance impact was *well*
lost in the noise.

This fixes https://github.com/git-for-windows/git/issues/501

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:53:45 -07:00
Nguyễn Thái Ngọc Duy
34e7771bc6 Use the right 'struct repository' instead of the_repository
There are a couple of places where 'struct repository' is already passed
around, but the_repository is still used. Use the right repo.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Nguyễn Thái Ngọc Duy
90d3405196 match-trees.c: remove the_repo from shift_tree*()
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Nguyễn Thái Ngọc Duy
0dd1f0c3a6 tree-walk.c: remove the_repo from get_tree_entry_follow_symlinks()
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Nguyễn Thái Ngọc Duy
50ddb089ff tree-walk.c: remove the_repo from get_tree_entry()
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Nguyễn Thái Ngọc Duy
5e57580733 tree-walk.c: remove the_repo from fill_tree_descriptor()
While at there, clean up the_repo usage in builtin/merge-tree.c a tiny
bit.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Nguyễn Thái Ngọc Duy
d3b4705ab8 sha1-file.c: remove the_repo from read_object_with_reference()
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 12:45:17 -07:00
Carlo Marcelo Arenas Belón
6114a405bd trace2: correct typo in technical documentation
an apparent typo for the environment variable was included with 81567caf87
("trace2: update docs to describe system/global config settings",
2019-04-15), and was missed when renamed variables by e4b75d6a1d
("trace2: rename environment variables to GIT_TRACE2*", 2019-05-19)

Signed-off-by: Carlo Marcelo Arenas Belón <carenas@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 10:56:54 -07:00
Phillip Wood
ed5b1ca10b status: do not report errors in sequencer/todo
commit 4a72486de9 ("fix cherry-pick/revert status after commit",
2019-04-16) used parse_insn_line() to parse the first line of the todo
list to check if it was a pick or revert. However if the todo list is
left over from an old cherry-pick or revert and references a commit that
no longer exists then parse_insn_line() prints an error message which is
confusing for users [1]. Instead parse just the command name so that the
user is alerted to the presence of stale sequencer state by status
reporting that a cherry-pick or revert is in progress.

Note that we should not be leaving stale sequencer state lying around
(or at least not as often) after commit b07d9bfd17 ("commit/reset: try
to clean up sequencer state", 2019-04-16). However the user may still
have stale state that predates that commit.

Also avoid printing an error message if for some reason the user has a
file called `sequencer` in $GIT_DIR.

[1] https://public-inbox.org/git/3bc58c33-4268-4e7c-bf1a-fe349b3cb037@www.fastmail.com/

Reported-by: Espen Antonsen <espen@inspired.no>
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-06-27 10:31:02 -07:00