Some codepaths failed to form a proper URL when .gitmodules record
the URL to a submodule repository as relative to the repository of
superproject, which has been corrected.
* sb/submodule-url-to-absolute:
submodule helper: convert relative URL to absolute URL if needed
"git repack" in a shallow clone did not correctly update the
shallow points in the repository, leading to a repository that
does not pass fsck.
* js/shallow-and-fetch-prune:
repack -ad: prune the list of shallow commits
shallow: offer to prune only non-existing entries
repack: point out a bug handling stale shallow info
The logic to determine the archive type "git archive" uses did not
correctly kick in for "git archive --remote", which has been
corrected.
* js/remote-archive-dwimfix:
archive: initialize archivers earlier
This helps format-patch gain completion for a couple new options,
notably --range-diff.
Since send-email completion relies on $__git_format_patch_options
which is now reduced, we need to do something not to regress
send-email completion.
The workaround here is implement --git-completion-helper in
send-email.perl just as a bridge to "format-patch --git-completion-helper".
This is enough to use __gitcomp_builtin on send-email (to take
advantage of caching).
In the end, send-email.perl can probably reuse the same info it passes
to GetOptions() to generate full --git-completion-helper output so
that we don't need to keep track of its options in git-completion.bash
anymore. But that's something for another boring day.
Helped-by: Denton Liu <liu.denton@gmail.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With refresh_index() learning to utilize preload_index() to speed up its
operation there is no longer any benefit to having the caller preload the
index first. Remove those unneeded calls by calling read_index() instead of
the preload variant.
There is no measurable performance impact of this patch - the 2nd call to
preload_index() bails out quickly but there is no reason to call it twice.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Normally pthread_create() rarely fails. But with new pthreads wrapper,
pthread_create() will return ENOSYS on a system without thread support.
Threaded code _is_ protected by HAVE_THREADS and pthread_create()
should never run in the first place. But the situation could change in
the future and bugs may sneak in. Make sure that all pthread_create()
reports the error cause.
While at there, mark these strings for translation if they aren't.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It was reported that when building with NO_PTHREADS=1,
-Wmaybe-uninitialized is triggered. Just initialize the variable from
the beginning to shut the compiler up (because this warning is enabled
in config.dev)
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is a faithful conversion with no attempt to clean up whatsoever.
Code indentation is left broken. There will be another commit to clean
it up and un-indent if we just indent now. It's just more code noise.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When NO_PTHREADS is still used in this file, we have two separate code
paths for thread and no thread support. The latter will always have
num_threads remain zero while the former uses num_threads zero as
"default number of threads".
With recent changes blur the line between thread and no-thread
support, this num_threads handling becomes a bit strange so let's
redefine it like this:
- num_threads == 0 means default number of threads and should become
positive after all configuration and option parsing is done if
multithread is supported.
- num_threads <= 1 runs no threads. It does not matter if the platform
supports threading or not.
- num_threads > 1 will run multiple threads and is invalid if
HAVE_THREADS is false. pthread API is only used in this case.
PS. a new warning is also added when num_threads is forced back to one
because a thread-incompatible option is specified.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is a faithful conversion without attempting to improve
anything. That comes later.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On systems that do not support multithread, start_async() is
implemented with fork(). This implementation details unfortunately
leak out at least in send-pack.c [1].
To keep the code base clean of NO_PTHREADS, move the this #ifdef back
to run-command.c. The new wrapper function async_with_fork() at least
helps suggest that this special "close()" is related to async in fork
mode.
[1] 09c9957cf7 (send-pack: avoid deadlock when pack-object dies early
- 2011-04-25)
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
run-command.c may use threads for its async support. But instead of
including pthread.h directly, let's include thread-utils.h.
run-command.c probably never needs the dummy bits in thread-utils.h
when NO_PTHREADS is defined. But this makes sure we have consistent
HAVE_THREADS behavior everywhere. From now on outside compat/,
thread-utils.h is the only place that includes pthread.h
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function was used only for parsing the hunk headers generated by
xdiff. Now that we can use hunk callbacks to get that information
directly, it has outlived its usefulness.
Note to anyone who wants to resurrect it: the "len" parameter was
totally unused, meaning that the function could read past the end of the
"line" array. In practice this never happened, because we only used it
to parse xdiff's generated header lines. But it would be dangerous to
use it for other cases without fixing this defect.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we count the lines in a diff, we don't actually care about the
contents of each line. By using a hunk callback, we tell xdiff that it
does not need to even bother generating a hunk header line, saving a
small amount of work.
Arguably we could even ignore the hunk headers completely, since we're
just computing a cost function between patches. But doing it this way
maintains the exact same behavior before and after.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "diff --check" code needs to know the line number on which each hunk
starts in order to generate its output. We get that now by parsing the
hunk header line generated by xdiff, but it's much simpler to just pass
it directly using a hunk callback.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A combined diff has to line up the hunks for all of the individual
pairwise diffs, and thus needs to know their line numbers and sizes. We
get that now by parsing the hunk header line that xdiff generates.
However, now that xdiff supports a hunk callback, we can just use the
values directly.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our word-diff does not look at the -/+ lines generated by xdiff at all
(because they are not real lines to show the user, but just the
tokenized words split into lines). Instead we use the line numbers from
the hunk headers to index our own data structure.
As a result, our xdi_diff_outf() callback throws away all lines except
hunk headers. We can instead use a hunk callback, which has two
benefits:
1. We don't have to re-parse the generated hunk header line, but can
use the passed parameters directly.
2. By setting our line callback to NULL, we can tell xdiff-interface
that it does not even need to bother generating the other lines,
saving a small amount of work.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do not include hunk header lines when computing patch-ids, since
the line numbers would create false negatives. Rather than detect and
skip them in our line callback, we can simply tell xdiff to avoid
generating them.
This is similar to the previous commit, but split out because it
actually requires modifying the matching line callback.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some callers of xdi_diff_outf() do not look at the generated hunk header
lines at all. By plugging in a no-op hunk callback, this tells xdiff not
to even bother formatting them.
This patch introduces a stock no-op callback and uses it with a few
callers whose line callbacks explicitly ignore hunk headers (because
they look only for +/- lines).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The value of timeout starts as an int value, and for this reason it
cannot overflow unsigned long long aka ULONGLONG. The unsigned version
of this initial value is available in orig_timeout. The difference
(orig_timeout - elapsed) cannot wrap around because it is protected by
a conditional (as can be seen in the patch text). Hence, the ULONGLONG
difference can only have values that are smaller than the initial
timeout value and truncation to int cannot overflow.
Signed-off-by: Steve Hoelzer <shoelzer@gmail.com>
[j6t: improved both implementation and log message]
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a dedicated signing subkey to the key identified as 'Eris
Discordia', and update tests appropriately. GnuPG will now sign commits
using the dedicated signing subkey, changing the value of %GK and %GF,
and effectively creating a test case for %GF!=%GP.
Signed-off-by: Michał Górny <mgorny@gentoo.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Test %GP in addition to %GF in custom format checks. With current
keyring, both have the same value.
Signed-off-by: Michał Górny <mgorny@gentoo.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
tree_entry_interesting() is used for matching pathspec on a tree. The
interesting thing about this function is that, because the tree
entries are known to be sorted, this function can return more than
just "yes, matched" and "no, not matched". It can also say "yes, this
entry is matched and so is the remaining entries in the tree".
This is where I made a mistake when matching exclude pathspec. For
exclude pathspec, we do matching twice, one with positive patterns and
one with negative ones, then a rule table is applied to determine the
final "include or exclude" result. Note that "matched" does not
necessarily mean include. For negative patterns, "matched" means
exclude.
This particular rule is too eager to include everything. Rule 8 says
that "if all entries are positively matched" and the current entry is
not negatively matched (i.e. not excluded), then all entries are
positively matched and therefore included. But this is not true. If
the _current_ entry is not negatively matched, it does not mean the
next one will not be and we cannot conclude right away that all
remaining entries are positively matched and can be included.
Rules 8 and 18 are now updated to be less eager. We conclude that the
current entry is positively matched and included. But we say nothing
about remaining entries. tree_entry_interesting() will be called again
for those entries where we will determine entries individually.
Reported-by: Christophe Bliard <christophe.bliard@trux.info>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Noticed-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Adjusting the build process to rely more on curl-config to populate
linker flags instead of manually populating flags based off detected
features.
Originally, a configure-invoked build would check for SSL-support in the
target curl library. If enabled, NEEDS_SSL_WITH_CURL would be set and
used in the Makefile to append additional libraries to link against. As
for systems building solely with make, the defines NEEDS_IDN_WITH_CURL
and NEEDS_SSL_WITH_CURL could be set to indirectly enable respective
linker flags. Since both configure.ac and Makefile already rely on
curl-config utility to provide curl-related build information, adjusting
the respective assets to populate required linker flags using the
utility (unless explicitly configured).
Signed-off-by: James Knight <james.d.knight@live.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On platforms with recent cURL library, http.sslBackend configuration
variable can be used to choose a different SSL backend at runtime.
The Windows port uses this mechanism to switch between OpenSSL and
Secure Channel while talking over the HTTPS protocol.
* js/mingw-http-ssl:
http: when using Secure Channel, ignore sslCAInfo by default
http: add support for disabling SSL revocation checks in cURL
http: add support for selecting SSL backends at runtime
New "--pretty=format:" placeholders %GF and %GP that show the GPG
key fingerprints have been invented.
* mg/gpg-fingerprint:
gpg-interface.c: obtain primary key fingerprint as well
gpg-interface.c: support getting key fingerprint via %GF format
gpg-interface.c: use flags to determine key/signer info presence
Detect and reject a signature block that has more than one GPG
signature.
* mg/gpg-parse-tighten:
gpg-interface.c: detect and reject multiple signatures on commits
Further clean-up of merge-recursive machinery.
* en/merge-cleanup-more:
merge-recursive: avoid showing conflicts with merge branch before HEAD
merge-recursive: improve auto-merging messages with path collisions
During an "add", a call is made to run_diff_files() which calls
check_removed() for each index-entry. The preload_index() code
distributes some of the costs across multiple threads.
Because the files checked are restricted to pathspec, adding
individual files makes no measurable impact but on a Windows repo
with ~200K files, 'git add .' drops from 6.3 seconds to 3.3 seconds
for a 47% savings.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The add_missing_tags() method currently has quadratic behavior.
This is due to a linear number (based on number of tags T) of
calls to in_merge_bases_many, which has linear performance (based
on number of commits C in the repository).
Replace this O(T * C) algorithm with an O(T + C) algorithm by
using get_reachable_subset(). We ignore the return list and focus
instead on the reachable_flag assigned to the commits we care
about, because we need to interact with the tag ref and not just
the commit object.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The get_reachable_subset() method returns the list of commits in
the 'to' array that are reachable from at least one commit in the
'from' array. Add tests that check this method works in a few
cases:
1. All commits in the 'to' list are reachable. This exercises the
early-termination condition.
2. Some commits in the 'to' list are reachable. This exercises the
loop-termination condition.
3. No commits in the 'to' list are reachable. This exercises the
NULL return condition.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The existing reachability algorithms in commit-reach.c focus on
finding merge-bases or determining if all commits in a set X can
reach at least one commit in a set Y. However, for two commits sets
X and Y, we may also care about which commits in Y are reachable
from at least one commit in X.
Implement get_reachable_subset() which answers this question. Given
two arrays of commits, 'from' and 'to', return a commit_list with
every commit from the 'to' array that is reachable from at least
one commit in the 'from' array.
The algorithm is a simple walk starting at the 'from' commits, using
the PARENT2 flag to indicate "this commit has already been added to
the walk queue". By marking the 'to' commits with the PARENT1 flag,
we can determine when we see a commit from the 'to' array. We remove
the PARENT1 flag as we add that commit to the result list to avoid
duplicates.
The order of the resulting list is a reverse of the order that the
commits are discovered in the walk.
There are a couple shortcuts to avoid walking more than we need:
1. We determine the minimum generation number of commits in the
'to' array. We do not walk commits with generation number
below this minimum.
2. We count how many distinct commits are in the 'to' array, and
decrement this count when we discover a 'to' commit during the
walk. If this number reaches zero, then we can terminate the
walk.
Tests will be added using the 'test-tool reach' helper in a
subsequent commit.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix a small bug introduced by "7a36987ff (send-email: add an auto option
for transfer encoding, 2018-07-14)".
I saw the following message when setting --transfer-encoding for a file
with the same encoding:
$ git send-email --transfer-encoding=8bit example.patch
Use of uninitialized value $xfer_encoding in concatenation (.) or string
at /usr/lib/git-core/git-send-email line 1744.
The new tests are by brian m. carlson.
Signed-off-by: Aaron Lindsay <aaron@aclindsay.com>
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The pathspec code always takes names to be matched as a
name/namelen pair, but match_attrs() never looks at namelen,
and just treats "name" like a NUL-terminated string, passing
it to git_check_attr().
This usually works anyway. Every caller passes a
NUL-terminated string, and in all but one the passed-in
length is the same as the length of the string (the
exception is dir_path_match(), which may pass a smaller
length to drop a trailing slash). So we won't currently ever
read random memory, and the one case I found actually
happens to work correctly because the attr code can handle
the trailing slash itself.
But it's still worth addressing, as the function interface
implies that the name does not have to be NUL-terminated,
making this an accident waiting to happen.
Since teaching git_check_attr() to take a ptr/len pair would
be a big refactor, we'll just allocate a new string. We can
do this only when necessary, which avoids paying the cost
for most callers.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The approxidate parser has a table of special keywords like
"yesterday", "noon", "pm", etc. Some of these, like "pm", do
the right thing if we've recently seen a number: "3pm" is
what you'd think.
However, most of them do not look at or modify the
pending-number flag at all, which means a number may "jump"
across a significant keyword and be used unexpectedly. For
example, when parsing:
January 5th noon pm
we'd connect the "5" to "pm", and ignore it as a
day-of-month. This is obviously a bit silly, as "noon"
already implies "pm". And other mis-parsed things are
generally as silly ("January 5th noon, years ago" would
connect the 5 to "years", but probably nobody would type
that).
However, the fix is simple: when we see a keyword like
"noon", we should flush the pending number (as we would if
we hit another number, or the end of the string). In a few
of the specials that actually modify the day, we can simply
throw away the number (saying "Jan 5 yesterday" should not
respect the number at all).
Note that we have to either move or forward-declare the
static pending_number() to make it accessible to these
functions; this patch moves it.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a traversal sees the --indexed-objects option, it adds
all blobs and valid cache-trees from the index to the
traversal using add_index_objects_to_pending(). But that
function totally ignores its flags parameter!
That means that doing:
git rev-list --objects --indexed-objects
and
git rev-list --objects --not --indexed-objects
produce the same output, because we ignore the UNINTERESTING
flag when walking the index in the second example.
Nobody noticed because this feature was added as a way for
tools like repack to increase their coverage of reachable
objects, meaning it would only be used like the first
example above.
But since it's user facing (and because the documentation
describes it "as if the objects are listed on the command
line"), we should make sure the negative case behaves
sensibly.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous commit taught xdiff to optionally provide the hunk header
data to a specialized callback. But most users of xdiff actually use our
more convenient xdi_diff_outf() helper, which ensures that our callbacks
are always fed whole lines.
Let's plumb the special hunk-callback through this interface, too. It
will follow the same rule as xdiff when the hunk callback is NULL (i.e.,
continue to pass a stringified hunk header to the line callback). Since
we add NULL to each caller, there should be no behavior change yet.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The xdiff library always emits hunk header lines to our callbacks as
formatted strings like "@@ -a,b +c,d @@\n". This is convenient if we're
going to output a diff, but less so if we actually need to compute using
those numbers, which requires re-parsing the line.
In preparation for moving away from this, let's teach xdiff a new
callback function which gets the broken-out hunk information. To help
callers that don't want to use this new callback, if it's NULL we'll
continue to format the hunk header into a string.
Note that this function renames the "outf" callback to "out_line", as
well. This isn't strictly necessary, but helps in two ways:
1. Now that there are two callbacks, it's nice to use more descriptive
names.
2. Many callers did not zero the emit_callback_data struct, and needed
to be modified to set ecb.out_hunk to NULL. By changing the name of
the existing struct member, that guarantees that any new callers
from in-flight topics will break the build and be examined
manually.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As we are working to rewrite some of the revision-walk machinery,
there could easily be some interesting interactions between the
options that force topological constraints (--topo-order,
--date-order, and --author-date-order) along with specifying a
path.
Add extra tests to t6012-rev-list-simplify.sh to add coverage of
these interactions. To ensure interesting things occur, alter the
repo data shape to have different orders depending on topo-, date-,
or author-date-order.
When testing using GIT_TEST_COMMIT_GRAPH, this assists in covering
the new logic for topo-order walks using generation numbers. The
extra tests can be added indepently.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current --topo-order algorithm requires walking all
reachable commits up front, topo-sorting them, all before
outputting the first value. This patch introduces a new
algorithm which uses stored generation numbers to
incrementally walk in topo-order, outputting commits as
we go. This can dramatically reduce the computation time
to write a fixed number of commits, such as when limiting
with "-n <N>" or filling the first page of a pager.
When running a command like 'git rev-list --topo-order HEAD',
Git performed the following steps:
1. Run limit_list(), which parses all reachable commits,
adds them to a linked list, and distributes UNINTERESTING
flags. If all unprocessed commits are UNINTERESTING, then
it may terminate without walking all reachable commits.
This does not occur if we do not specify UNINTERESTING
commits.
2. Run sort_in_topological_order(), which is an implementation
of Kahn's algorithm. It first iterates through the entire
set of important commits and computes the in-degree of each
(plus one, as we use 'zero' as a special value here). Then,
we walk the commits in priority order, adding them to the
priority queue if and only if their in-degree is one. As
we remove commits from this priority queue, we decrement the
in-degree of their parents.
3. While we are peeling commits for output, get_revision_1()
uses pop_commit on the full list of commits computed by
sort_in_topological_order().
In the new algorithm, these three steps correspond to three
different commit walks. We run these walks simultaneously,
and advance each only as far as necessary to satisfy the
requirements of the 'higher order' walk. We know when we can
pause each walk by using generation numbers from the commit-
graph feature.
Recall that the generation number of a commit satisfies:
* If the commit has at least one parent, then the generation
number is one more than the maximum generation number among
its parents.
* If the commit has no parent, then the generation number is one.
There are two special generation numbers:
* GENERATION_NUMBER_INFINITY: this value is 0xffffffff and
indicates that the commit is not stored in the commit-graph and
the generation number was not previously calculated.
* GENERATION_NUMBER_ZERO: this value (0) is a special indicator
to say that the commit-graph was generated by a version of Git
that does not compute generation numbers (such as v2.18.0).
Since we use generation_numbers_enabled() before using the new
algorithm, we do not need to worry about GENERATION_NUMBER_ZERO.
However, the existence of GENERATION_NUMBER_INFINITY implies the
following weaker statement than the usual we expect from
generation numbers:
If A and B are commits with generation numbers gen(A) and
gen(B) and gen(A) < gen(B), then A cannot reach B.
Thus, we will walk in each of our stages until the "maximum
unexpanded generation number" is strictly lower than the
generation number of a commit we are about to use.
The walks are as follows:
1. EXPLORE: using the explore_queue priority queue (ordered by
maximizing the generation number), parse each reachable
commit until all commits in the queue have generation
number strictly lower than needed. During this walk, update
the UNINTERESTING flags as necessary.
2. INDEGREE: using the indegree_queue priority queue (ordered
by maximizing the generation number), add one to the in-
degree of each parent for each commit that is walked. Since
we walk in order of decreasing generation number, we know
that discovering an in-degree value of 0 means the value for
that commit was not initialized, so should be initialized to
two. (Recall that in-degree value "1" is what we use to say a
commit is ready for output.) As we iterate the parents of a
commit during this walk, ensure the EXPLORE walk has walked
beyond their generation numbers.
3. TOPO: using the topo_queue priority queue (ordered based on
the sort_order given, which could be commit-date, author-
date, or typical topo-order which treats the queue as a LIFO
stack), remove a commit from the queue and decrement the
in-degree of each parent. If a parent has an in-degree of
one, then we add it to the topo_queue. Before we decrement
the in-degree, however, ensure the INDEGREE walk has walked
beyond that generation number.
The implementations of these walks are in the following methods:
* explore_walk_step and explore_to_depth
* indegree_walk_step and compute_indegrees_to_depth
* next_topo_commit and expand_topo_walk
These methods have some patterns that may seem strange at first,
but they are probably carry-overs from their equivalents in
limit_list and sort_in_topological_order.
One thing that is missing from this implementation is a proper
way to stop walking when the entire queue is UNINTERESTING, so
this implementation is not enabled by comparisions, such as in
'git rev-list --topo-order A..B'. This can be updated in the
future.
In my local testing, I used the following Git commands on the
Linux repository in three modes: HEAD~1 with no commit-graph,
HEAD~1 with a commit-graph, and HEAD with a commit-graph. This
allows comparing the benefits we get from parsing commits from
the commit-graph and then again the benefits we get by
restricting the set of commits we walk.
Test: git rev-list --topo-order -100 HEAD
HEAD~1, no commit-graph: 6.80 s
HEAD~1, w/ commit-graph: 0.77 s
HEAD, w/ commit-graph: 0.02 s
Test: git rev-list --topo-order -100 HEAD -- tools
HEAD~1, no commit-graph: 9.63 s
HEAD~1, w/ commit-graph: 6.06 s
HEAD, w/ commit-graph: 0.06 s
This speedup is due to a few things. First, the new generation-
number-enabled algorithm walks commits on order of the number of
results output (subject to some branching structure expectations).
Since we limit to 100 results, we are running a query similar to
filling a single page of results. Second, when specifying a path,
we must parse the root tree object for each commit we walk. The
previous benefits from the commit-graph are entirely from reading
the commit-graph instead of parsing commits. Since we need to
parse trees for the same number of commits as before, we slow
down significantly from the non-path-based query.
For the test above, I specifically selected a path that is changed
frequently, including by merge commits. A less-frequently-changed
path (such as 'README') has similar end-to-end time since we need
to walk the same number of commits (before determining we do not
have 100 hits). However, get the benefit that the output is
presented to the user as it is discovered, much the same as a
normal 'git log' command (no '--topo-order'). This is an improved
user experience, even if the command has the same runtime.
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are a few things that need to move around a little before
making a big refactoring in the topo-order logic:
1. We need access to record_author_date() and
compare_commits_by_author_date() in revision.c. These are used
currently by sort_in_topological_order() in commit.c.
2. Moving these methods to commit.h requires adding an author_date_slab
declaration to commit.h. Consumers will need their own implementation.
3. The add_parents_to_list() method in revision.c performs logic
around the UNINTERESTING flag and other special cases depending
on the struct rev_info. Allow this method to ignore a NULL 'list'
parameter, as we will not be populating the list for our walk.
Also rename the method to the slightly more generic name
process_parents() to make clear that this method does more than
add to a list (and no list is required anymore).
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
