Jonathan Tan [Wed, 21 Aug 2019 22:20:10 +0000 (15:20 -0700)]
transport: teach all vtables to allow fetch first
The only transport that does not allow fetch() to be called before
get_refs_list() is the bundle transport. Clean up the code by teaching
the bundle transport the ability to do this, and removing support for
transports that don't support this order of invocation.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jonathan Tan [Wed, 21 Aug 2019 22:20:09 +0000 (15:20 -0700)]
transport-helper: skip ls-refs if unnecessary
Commit e70a3030e7 ("fetch: do not list refs if fetching only hashes",
2018-10-07) and its ancestors taught Git, as an optimization, to skip
the ls-refs step when it is not necessary during a protocol v2 fetch
(for example, when lazy fetching a missing object in a partial clone, or
when running "git fetch --no-tags <remote> <SHA-1>"). But that was only
done for natively supported protocols; in particular, HTTP was not
supported.
Teach Git to skip ls-refs when using remote helpers that support connect
or stateless-connect. To do this, fetch() is made an acceptable entry
point. Because fetch() can now be the first function in the vtable
called, "get_helper(transport);" has to be added to the beginning of
that function to set the transport up (if not yet set up) before
process_connect() is invoked.
When fetch() is called, the transport could be taken over (this happens
if "connect" or "stateless-connect" is successfully run without any
"fallback" response), or not. If the transport is taken over, execution
continues like execution for natively supported protocols
(fetch_refs_via_pack() is executed, which will fetch refs using ls-refs
if needed). If not, the remote helper interface will invoke
get_refs_list() if it hasn't been invoked yet, preserving existing
behavior.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano [Thu, 22 Aug 2019 19:34:11 +0000 (12:34 -0700)]
Merge branch 'sg/show-failed-test-names'
The first line of verbose output from each test piece now carries
the test name and number to help scanning with eyeballs.
* sg/show-failed-test-names:
tests: show the test name and number at the start of verbose output
t0000-basic: use realistic test script names in the verbose tests
Junio C Hamano [Thu, 22 Aug 2019 19:34:11 +0000 (12:34 -0700)]
Merge branch 'sg/commit-graph-validate'
The code to write commit-graph over given commit object names has
been made a bit more robust.
* sg/commit-graph-validate:
commit-graph: error out on invalid commit oids in 'write --stdin-commits'
commit-graph: turn a group of write-related macro flags into an enum
t5318-commit-graph: use 'test_expect_code'
"git checkout" and "git restore" to re-populate the index from a
tree-ish (typically HEAD) did not work correctly for a path that
was removed and then added again with the intent-to-add bit, when
the corresponding working tree file was empty. This has been
corrected.
* vn/restore-empty-ita-corner-case-fix:
restore: add test for deleted ita files
checkout.c: unstage empty deleted ita files
Junio C Hamano [Thu, 22 Aug 2019 19:34:10 +0000 (12:34 -0700)]
Merge branch 'sg/do-not-skip-non-httpd-tests'
Test fix.
* sg/do-not-skip-non-httpd-tests:
t: warn against adding non-httpd-specific tests after sourcing 'lib-httpd'
t5703: run all non-httpd-specific tests before sourcing 'lib-httpd.sh'
t5510-fetch: run non-httpd-specific test before sourcing 'lib-httpd.sh'
Junio C Hamano [Thu, 22 Aug 2019 19:34:10 +0000 (12:34 -0700)]
Merge branch 'jk/tree-walk-overflow'
Codepaths to walk tree objects have been audited for integer
overflows and hardened.
* jk/tree-walk-overflow:
tree-walk: harden make_traverse_path() length computations
tree-walk: add a strbuf wrapper for make_traverse_path()
tree-walk: accept a raw length for traverse_path_len()
tree-walk: use size_t consistently
tree-walk: drop oid from traverse_info
setup_traverse_info(): stop copying oid
Junio C Hamano [Thu, 22 Aug 2019 19:34:10 +0000 (12:34 -0700)]
Merge branch 'mt/grep-submodules-working-tree'
"git grep --recurse-submodules" that looks at the working tree
files looked at the contents in the index in submodules, instead of
files in the working tree.
* mt/grep-submodules-working-tree:
grep: fix worktree case in submodules
Thomas Gummerer [Thu, 22 Aug 2019 19:22:40 +0000 (20:22 +0100)]
t0021: make sure clean filter runs
In t0021.15 one of the things we are checking is that the clean filter
is run when checking out empty-branch. The clean filter needs to be
run to make sure there are no modifications on the file system for the
test.r file, and thus it isn't dangerous to overwrite it.
However in the current test setup it is not always necessary to run
the clean filter, and thus the test sometimes fails, as debug.log
isn't written.
This happens when test.r has an older mtime than the index itself.
That mtime is also recorded as stat data for test.r in the index, and
based on the heuristic we're using for index entries, git correctly
assumes this file is up-to-date.
Usually this test succeeds because the mtime of test.r is the same as
the mtime of the index. In this case test.r is racily clean, so git
actually checks the contents, for which the clean filter is run.
Fix the test by updating the mtime of test.r, so git is forced to
check the contents of the file, and the clean filter is run as the
test expects.
Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Mischa POSLAWSKY [Sat, 17 Aug 2019 21:51:07 +0000 (23:51 +0200)]
ref-filter: initialize empty name or email fields
Formatting $(taggername) on headerless tags such as v0.99 in Git
causes a SIGABRT with error "munmap_chunk(): invalid pointer",
because of an oversight in commit f0062d3b74 (ref-filter: free
item->value and item->value->s, 2018-10-19).
Signed-off-by: Mischa POSLAWSKY <git@shiar.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Stephen Boyd [Mon, 19 Aug 2019 21:22:43 +0000 (14:22 -0700)]
userdiff: add a builtin pattern for dts files
The Linux kernel receives many patches to the devicetree files each
release. The hunk header for those patches typically show nothing,
making it difficult to figure out what node is being modified without
applying the patch or opening the file and seeking to the context. Let's
add a builtin 'dts' pattern to git so that users can get better diff
output on dts files when they use the diff=dts driver.
The regex has been constructed based on the spec at devicetree.org[1]
and with some help from Johannes Sixt.
Cc: Rob Herring <robh+dt@kernel.org> Cc: Frank Rowand <frowand.list@gmail.com> Signed-off-by: Stephen Boyd <sboyd@kernel.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
SZEDER Gábor [Wed, 21 Aug 2019 11:04:24 +0000 (13:04 +0200)]
line-log: avoid unnecessary full tree diffs
With rename detection enabled the line-level log is able to trace the
evolution of line ranges across whole-file renames [1]. Alas, to
achieve that it uses the diff machinery very inefficiently, making the
operation very slow [2]. And since rename detection is enabled by
default, the line-level log is very slow by default.
When the line-level log processes a commit with rename detection
enabled, it currently does the following (see queue_diffs()):
1. Computes a full tree diff between the commit and (one of) its
parent(s), i.e. invokes diff_tree_oid() with an empty
'diffopt->pathspec'.
2. Checks whether any paths in the line ranges were modified.
3. Checks whether any modified paths in the line ranges are missing
in the parent commit's tree.
4. If there is such a missing path, then calls diffcore_std() to
figure out whether the path was indeed renamed based on the
previously computed full tree diff.
5. Continues doing stuff that are unrelated to the slowness.
So basically the line-level log computes a full tree diff for each
commit-parent pair in step (1) to be used for rename detection in step
(4) in the off chance that an interesting path is missing from the
parent.
Avoid these expensive and mostly unnecessary full tree diffs by
limiting the diffs to paths in the line ranges. This is much cheaper,
and makes step (2) unnecessary. If it turns out that an interesting
path is missing from the parent, then fall back and compute a full
tree diff, so the rename detection will still work.
Care must be taken when to update the pathspec used to limit the diff
in case of renames. A path might be renamed on one branch and
modified on several parallel running branches, and while processing
commits on these branches the line-level log might have to alternate
between looking at a path's new and old name. However, at any one
time there is only a single 'diffopt->pathspec'.
So add a step (0) to the above to ensure that the paths in the
pathspec match the paths in the line ranges associated with the
currently processed commit, and re-parse the pathspec from the paths
in the line ranges if they differ.
The new test cases include a specially crafted piece of history with
two merged branches and two files, where each branch modifies both
files, renames on of them, and then modifies both again. Then two
separate 'git log -L' invocations check the line-level log of each of
those two files, which ensures that at least one of those invocations
have to do that back-and-forth between the file's old and new name (no
matter which branch is traversed first). 't/t4211-line-log.sh'
already contains two tests involving renames, they don't don't trigger
this back-and-forth.
Avoiding these unnecessary full tree diffs can have huge impact on
performance, especially in big repositories with big trees and mergy
history. Tracing the evolution of a function through the whole
history:
# git.git
$ time git --no-pager log -L:read_alternate_refs:sha1-file.c v2.23.0
Before:
real 0m8.874s
user 0m8.816s
sys 0m0.057s
After:
real 0m2.516s
user 0m2.456s
sys 0m0.060s
# linux.git
$ time ~/src/git/git --no-pager log \
-L:build_restore_work_registers:arch/mips/mm/tlbex.c v5.2
Before:
real 3m50.033s
user 3m48.041s
sys 0m0.300s
After:
real 0m2.599s
user 0m2.466s
sys 0m0.157s
That's just over 88x speedup.
[1] Line-level log's rename following is quite similar to 'git log
--follow path', with the notable differences that it does handle
multiple paths at once as well, and that it doesn't show the
commit performing the rename if it's an exact rename.
[2] This slowness might not have been apparent initially, because back
when the line-level log feature was introduced rename detection
was not yet enabled by default; 12da1d1f6f (Implement line-history
search (git log -L), 2013-03-28) and 5404c116aa (diff: activate
diff.renames by default, 2016-02-25).
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jonathan Tan [Tue, 20 Aug 2019 20:53:20 +0000 (13:53 -0700)]
diff: skip GITLINK when lazy fetching missing objs
In 7fbbcb21b1 ("diff: batch fetching of missing blobs", 2019-04-08),
diff was taught to batch the fetching of missing objects when operating
on a partial clone, but was not taught to refrain from fetching
GITLINKs. Teach diff to check if an object is a GITLINK before including
it in the set to be fetched.
(As stated in the commit message of that commit, unpack-trees was also
taught a similar thing prior, but unpack-trees correctly checks for
GITLINK before including objects in the set to be fetched.)
Denton Liu [Mon, 19 Aug 2019 06:26:19 +0000 (02:26 -0400)]
packfile.h: drop extern from function declaration
In 336226c259 (packfile.h: drop extern from function declarations,
2019-04-05), `extern` was removed from function declarations because
it's redundant. However, in 8434e85d5f (repack: refactor pack deletion
for future use, 2019-06-10), an `extern` was mistakenly included.
Remove this spurious `extern`.
Signed-off-by: Denton Liu <liu.denton@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:42 +0000 (19:16 +0000)]
t3800: make hash-size independent
Replace references to several hard-coded object IDs with a variable
referring to the generated commit. Avoid matching on exact character
positions, which will be different depending on the hash in use. In the
test for a valid object ID, use an obviously invalid one from the lookup
table.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:40 +0000 (19:16 +0000)]
t3506: make hash independent
This test uses a hard-coded object ID to ensure that the result of
cherry-pick --ff is correct. Use test_oid to make this work for both
SHA-1 and SHA-256.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:38 +0000 (19:16 +0000)]
t3404: abstract away SHA-1-specific constants
Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes. Add a use of $EMPTY_TREE instead of a
hard-coded value. Remove a comment about hard-coded hashes which is no
longer applicable.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:37 +0000 (19:16 +0000)]
t3306: abstract away SHA-1-specific constants
Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes. Convert some single-line heredocs into inline
uses of echo now that they can be expressed succinctly.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:36 +0000 (19:16 +0000)]
t3305: make hash size independent
Instead of hard-coding 40-character shell patterns, use grep to
determine if all of the paths have either zero or one levels of fanout,
as appropriate.
Note that the final test is implicitly dependent on the hash algorithm.
Depending on the algorithm in use, the fanout may or may not completely
compress. In its current state, this is not a problem, but it could be
if the hash algorithm changes again.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 19:16:35 +0000 (19:16 +0000)]
t3301: abstract away SHA-1-specific constants
Adjust the test so that it computes variables for object IDs instead of
using hard-coded hashes. Move some invocations of test_commit around so
that we can compute the object IDs for these commits.
Compute several object IDs in the tests instead of using hard-coded
values so that the test works with any hash algorithm. Since the actual
values are sorted by the object ID of the object being annotated, sort
the expected values accordingly as well.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Phillip Wood [Mon, 19 Aug 2019 09:18:23 +0000 (02:18 -0700)]
sequencer: simplify root commit creation
Adapt try_to_commit() to create a new root commit rather than special
casing this in run_git_commit(). This significantly reduces the amount of
special case code for creating the root commit and reduces the number of
commit code paths we have to worry about.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Phillip Wood [Mon, 19 Aug 2019 09:18:22 +0000 (02:18 -0700)]
rebase -i: check for updated todo after squash and reword
While a rebase is stopped for the user to edit a commit message it can
be convenient for them to also edit the todo list. The scripted version
of rebase supported this but the C version does not. We already check to
see if the todo list has been updated by an exec command so extend this
to rewords and squashes. It only costs a single stat call to do this so
it should not affect the speed of the rebase (especially as it has just
stopped for the user to edit a message)
Note that for squashes the editor may be opened on a different pick to
the squash itself as we edit the message at the end of a chain fixups
and squashes.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Phillip Wood [Mon, 19 Aug 2019 09:18:21 +0000 (02:18 -0700)]
rebase -i: always update HEAD before rewording
If the user runs git log while rewording a commit it is confusing if
sometimes we're amending the commit that's being reworded and at other
times we're creating a new commit depending on whether we could
fast-forward or not[1]. Fix this inconsistency by always committing the
picked commit and then running 'git commit --amend' to do the reword.
The first commit is performed by the sequencer without forking git
commit and does not impact on the speed of rebase. In a test rewording
100 commits with
GIT_EDITOR=true GIT_SEQUENCE_EDITOR='sed -i s/pick/reword/' \
../bin-wrappers/git rebase -i --root
and taking the best of three runs the current master took
957ms and with this patch it took 961ms.
This change fixes rewording the new root commit when rearranging commits
with --root.
Note that the new code no longer updates CHERRY_PICK_HEAD after creating
a root commit - I'm not sure why the old code was that creating that ref
after a successful commit, everywhere else it is removed after a
successful commit.
Reported-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Beat Bolli [Sun, 18 Aug 2019 20:17:27 +0000 (22:17 +0200)]
grep: under --debug, show whether PCRE JIT is enabled
This information is useful and not visible anywhere else, so show it.
Signed-off-by: Beat Bolli <dev+git@drbeat.li> Suggested-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:27 +0000 (20:04 +0000)]
midx: switch to using the_hash_algo
Instead of hard-coding the hash size, use the_hash_algo to look up the
hash size at runtime. Remove the #define constant which was used to
hold the hash length, since writing the expression with the_hash_algo
provide enough documentary value on its own.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:26 +0000 (20:04 +0000)]
builtin/show-index: replace sha1_to_hex
In this code path, we use sha1_to_hex to display the contents of a v1
pack index. While we plan to switch to v3 indices for SHA-256, the v1
pack indices still function, so to support both algorithms, switch
sha1_to_hex to hash_to_hex.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:25 +0000 (20:04 +0000)]
rerere: replace sha1_to_hex
Replace the uses of sha1_to_hex in this function with hash_to_hex to
allow the use of SHA-256 as well. Rename a variable since it is no
longer limited to SHA-1.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:20 +0000 (20:04 +0000)]
cache: remove null_sha1
All of the existing uses of null_sha1 can be converted into uses of
null_oid, so do so. Remove null_sha1 and is_null_sha1, and define
is_null_oid in terms of null_oid. This also has the additional benefit
of removing several uses of sha1_to_hex.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:09 +0000 (20:04 +0000)]
show-index: switch hard-coded constants to the_hash_algo
show-index uses a variety of hard-coded constants to enumerate the
values in pack indices. Switch these hard-coded constants to use
the_hash_algo or GIT_MAX_RAWSZ, as appropriate, and convert one instance
of an unsigned char array to struct object_id.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:08 +0000 (20:04 +0000)]
blame: remove needless comparison with GIT_SHA1_HEXSZ
When faking a working tree commit, we read in lines from MERGE_HEAD into
a strbuf. Because the strbuf is NUL-terminated and get_oid_hex will
fail if it unexpectedly encounters a NUL, the check for the length of
the line is unnecessary. There is no optimization benefit from this
case, either, since on failure we call die. Remove this check, since it
is no longer needed.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
brian m. carlson [Sun, 18 Aug 2019 20:04:05 +0000 (20:04 +0000)]
builtin/receive-pack: switch to use the_hash_algo
The push cert code uses HMAC-SHA-1 to create a nonce. This is a secure
use of SHA-1 which is not affected by its collision resistance (or lack
thereof). However, it makes sense for us to use a better algorithm if
one is available, one which may even be more performant. Futhermore,
until we have specialized functions for computing the hex value of an
arbitrary function, it simplifies the code greatly to use the same hash
algorithm everywhere.
Switch this code to use GIT_MAX_BLKSZ and the_hash_algo for computing
the push cert nonce, and rename the hmac_sha1 function to simply "hmac".
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Corentin BOMPARD [Mon, 19 Aug 2019 09:11:20 +0000 (11:11 +0200)]
pull, fetch: add --set-upstream option
Add the --set-upstream option to git pull/fetch
which lets the user set the upstream configuration
(branch.<current-branch-name>.merge and
branch.<current-branch-name>.remote) for the current branch.
A typical use-case is:
git clone http://example.com/my-public-fork
git remote add main http://example.com/project-main-repo
git pull --set-upstream main master
or, instead of the last line:
git fetch --set-upstream main master
git merge # or git rebase
This is mostly equivalent to cloning project-main-repo (which sets
upsteam) and then "git remote add" my-public-fork, but may feel more
natural for people using a hosting system which allows forking from
the web UI.
This functionality is analog to "git push --set-upstream".
Signed-off-by: Corentin BOMPARD <corentin.bompard@etu.univ-lyon1.fr> Signed-off-by: Nathan BERBEZIER <nathan.berbezier@etu.univ-lyon1.fr> Signed-off-by: Pablo CHABANNE <pablo.chabanne@etu.univ-lyon1.fr> Signed-off-by: Matthieu Moy <git@matthieu-moy.fr> Patch-edited-by: Matthieu Moy <git@matthieu-moy.fr> Signed-off-by: Junio C Hamano <gitster@pobox.com>
René Scharfe [Sat, 17 Aug 2019 16:24:23 +0000 (18:24 +0200)]
archive-tar: turn length miscalculation warning into BUG
Now that we're confident our pax extended header calculation is correct,
turn the criticality of the assertion up to the maximum, from warning
right up to BUG. Simplify the test, as the stderr comparison step would
not be reached in case the BUG message is triggered.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
A pax extended header record starts with a decimal number. Its value
is the length of the whole record, including its own length.
The calculation of that number in strbuf_append_ext_header() is off by
one in case the length of the rest is close to a higher order of
magnitude. This affects paths and link targets a bit shorter than 1000,
10000, 100000 etc. characters -- paths with a length of up to 100 fit
into the tar header and don't need a pax extended header.
The mistake has been present since the function was added by ae64bbc18c
("tar-tree: Introduce write_entry()", 2006-03-25).
Account for digits added to len during the loop and keep incrementing
until we have enough space for len and the rest. The crucial change is
to check against the current value of len before each iteration, instead
of against its value before the loop.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extended header entries contain a length value that is a bit tricky to
calculate because it includes its own length (number of decimal digits)
as well. We get it wrong in corner cases. Add a check, report wrong
results as a warning and add a test for exercising it.
Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:44 +0000 (11:41 -0700)]
merge-recursive: alphabetize include list
Other than cache.h which needs to appear first, and merge-recursive.h
which I want to be second so that we are more likely to notice if
merge-recursive.h has any missing includes, the rest of the list is
long and easier to look through if it's alphabetical.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:43 +0000 (11:41 -0700)]
merge-recursive: add sanity checks for relevant merge_options
There are lots of options that callers can set, yet most have a limited
range of valid values, some options are meant for output (e.g.
opt->obuf, which is expected to start empty), and callers are expected
to not set opt->priv. Add several sanity checks to ensure callers
provide sane values.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:42 +0000 (11:41 -0700)]
merge-recursive: rename MERGE_RECURSIVE_* to MERGE_VARIANT_*
I want to implement the same outward facing API as found within
merge-recursive.h in a different merge strategy. However, that makes
names like MERGE_RECURSIVE_{NORMAL,OURS,THEIRS} look a little funny;
rename to MERGE_VARIANT_{NORMAL,OURS,THEIRS}.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:41 +0000 (11:41 -0700)]
merge-recursive: split internal fields into a separate struct
merge_options has several internal fields that should not be set or read
by external callers. This just complicates the API. Move them into an
opaque merge_options_internal struct that is defined only in
merge-recursive.c and keep these out of merge-recursive.h.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:40 +0000 (11:41 -0700)]
merge-recursive: avoid losing output and leaking memory holding that output
If opt->buffer_output is less than 2, then merge_trees(),
merge_recursive(), and merge_recursive_generic() are all supposed to
flush the opt->obuf output buffer to stdout and release any memory it
holds. merge_trees() did not do this. Move the logic that handles this
for merge_recursive_internal() to merge_finalize() so that all three
methods handle this requirement.
Note that this bug didn't cause any problems currently, because there
are only two callers of merge_trees() right now (a git grep for
'merge_trees(' is misleading because builtin/merge-tree.c also defines a
'merge_tree' function that is unrelated), and only one of those is
called with buffer_output less than 2 (builtin/checkout.c), but it set
opt->verbosity to 0, for which there is only currently one non-error
message that would be shown: "Already up to date!". However, that one
message can only occur when the merge is utterly trivial (the merge base
tree exactly matches the merge tree), and builtin/checkout.c already
attempts a trivial merge via unpack_trees() before falling back to
merge_trees().
Also, if opt->buffer_output is 2, then the caller is responsible to
handle showing any output in opt->obuf and for free'ing it. This
requirement might be easy to overlook, so add a comment to
merge-recursive.h pointing it out. (There are currently two callers
that set buffer_output to 2, both in sequencer.c, and both of which
handle this correctly.)
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:39 +0000 (11:41 -0700)]
merge-recursive: comment and reorder the merge_options fields
The merge_options struct had lots of fields, making it a little
imposing, but the options naturally fall into multiple different groups.
Grouping similar options and adding a comment or two makes it easier to
read, easier for new folks to figure out which options are related, and
thus easier for them to find the options they need.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:38 +0000 (11:41 -0700)]
merge-recursive: consolidate unnecessary fields in merge_options
We provided users with the ability to state whether they wanted rename
detection, and to put a limit on how much CPU would be spent. Both of
these fields had multiple configuration parameters for setting them,
with one being a fallback and the other being an override. However,
instead of implementing the logic for how to combine the multiple
source locations into the appropriate setting at config loading time,
we loaded and tracked both values and then made the code combine them
every time it wanted to check the overall value. This had a few
minor drawbacks:
* it seems more complicated than necessary
* it runs the risk of people using the independent settings in the
future and breaking the intent of how the options are used
together
* it makes merge_options more complicated than necessary for other
potential users of the API
Fix these problems by moving the logic for combining the pairs of
options into a single value; make it apply at time-of-config-loading
instead of each-time-of-use.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:37 +0000 (11:41 -0700)]
merge-recursive: move some definitions around to clean up the header
No substantive code changes (view this with diff --color-moved), but
a few small code cleanups:
* Move structs and an inline function only used by merge-recursive.c
into merge-recursive.c
* Re-order function declarations to be more logical
* Add or fix some explanatory comments
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:36 +0000 (11:41 -0700)]
merge-recursive: rename merge_options argument to opt in header
In commit 259ccb6cc324 ("merge-recursive: rename merge_options argument
from 'o' to 'opt'", 2019-04-05), I renamed a bunch of function
arguments in merge-recursive.c, but forgot to make that same change to
merge-recursive.h. Make the two match.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:34 +0000 (11:41 -0700)]
merge-recursive: use common name for ancestors/common/base_list
merge_trees(), merge_recursive(), and merge_recursive_generic() in
their function headers used four different names for the merge base or
list of merge bases they were passed:
* 'common'
* 'ancestors'
* 'ca'
* 'base_list'
They were able to refer to it four different ways instead of only three
by using a different name in the signature for the .c file than the .h
file. Change all of these to 'merge_base' or 'merge_bases'.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:33 +0000 (11:41 -0700)]
merge-recursive: fix some overly long lines
No substantive code change, just add some line breaks to fix lines that
have grown in length due to various refactorings. Most remaining lines
of excessive length in merge-recursive include error messages and it's
not clear that splitting those improves things.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:32 +0000 (11:41 -0700)]
cache-tree: share code between functions writing an index as a tree
write_tree_from_memory() appeared to be a merge-recursive special that
basically duplicated write_index_as_tree(). The two have a different
signature, but the bigger difference was just that write_index_as_tree()
would always unconditionally read the index off of disk instead of
working on the current in-memory index. So:
* split out common code into write_index_as_tree_internal()
* rename write_tree_from_memory() to write_inmemory_index_as_tree(),
make it call write_index_as_tree_internal(), and move it to
cache-tree.c
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:31 +0000 (11:41 -0700)]
merge-recursive: don't force external callers to do our logging
Alternatively, you can view this as "make the merge functions behave
more similarly." merge-recursive has three different entry points:
merge_trees(), merge_recursive(), and merge_recursive_generic(). Two of
these would call diff_warn_rename_limit(), but merge_trees() didn't.
This lead to callers of merge_trees() needing to manually call
diff_warn_rename_limit() themselves. Move this to the new
merge_finalize() function to make sure that all three entry points run
this function.
Note that there are two external callers of merge_trees(), one in
sequencer.c and one in builtin/checkout.c. The one in sequencer.c is
cleaned up by this patch and just transfers where the call to
diff_warn_rename_limit() is made; the one in builtin/checkout.c is for
switching to a different commit and in the very rare case where the
warning might be triggered, it would probably be helpful to include
(e.g. if someone is modifying a file that has been renamed in moving to
the other commit, but there are so many renames between the commits that
the limit kicks in and none are detected, it may help to have an
explanation about why they got a delete/modify conflict instead of a
proper content merge in a renamed file).
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:30 +0000 (11:41 -0700)]
merge-recursive: remove useless parameter in merge_trees()
merge_trees() took a results parameter that would only be written when
opt->call_depth was positive, which is never the case now that
merge_trees_internal() has been split from merge_trees(). Remove the
misleading and unused parameter from merge_trees().
While at it, add some comments explaining how the output of
merge_trees() and merge_recursive() differ.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:29 +0000 (11:41 -0700)]
merge-recursive: exit early if index != head
We had a rule to enforce that the index matches head, but it was found
at the beginning of merge_trees() and would only trigger when
opt->call_depth was 0. Since merge_recursive() doesn't call
merge_trees() until after returning from recursing, this meant that the
check wasn't triggered by merge_recursive() until it had first finished
all the intermediate merges to create virtual merge bases. That is a
potentially huge amount of computation (and writing of intermediate
merge results into the .git/objects directory) before it errors out and
says, in effect, "Sorry, I can't do any merging because you have some
local changes that would be overwritten."
Further, not enforcing this requirement earlier allowed other bugs (such
as an unintentional unconditional dropping and reloading of the index in
merge_recursive() even when no recursion was necessary), to mask bugs in
other callers (which were fixed in the commit prior to this one).
Make sure we do the index == head check at the beginning of the merge,
and error out immediately if it fails. While we're at it, fix a small
leak in the show-the-error codepath.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:28 +0000 (11:41 -0700)]
Ensure index matches head before invoking merge machinery, round N
This is the bug that just won't die; there always seems to be another
form of it somewhere. See the commit message of 55f39cf7551b ("merge:
fix misleading pre-merge check documentation", 2018-06-30) for a more
detailed explanation), but in short:
<quick summary>
builtin/merge.c contains this important requirement for merge
strategies:
...the index must be in sync with the head commit. The strategies are
responsible to ensure this.
This condition is important to enforce because there are two likely
failure cases when the index isn't in sync with the head commit:
* we silently throw away changes the user had staged before the merge
* we accidentally (and silently) include changes in the merge that
were not part of either of the branches/trees being merged
Discarding users' work and mis-merging are both bad outcomes, especially
when done silently, so naturally this rule was stated sternly -- but,
unfortunately totally ignored in practice unless and until actual bugs
were found. But, fear not: the bugs from this were fixed in commit ee6566e8d70d ("[PATCH] Rewrite read-tree", 2005-09-05)
through a rewrite of read-tree (again, commit 55f39cf7551b has a more
detailed explanation of how this affected merge). And it was fixed
again in commit 160252f81626 ("git-merge-ours: make sure our index matches HEAD", 2005-11-03)
...and it was fixed again in commit 3ec62ad9ffba ("merge-octopus: abort if index does not match HEAD", 2016-04-09)
...and again in commit 65170c07d466 ("merge-recursive: avoid incorporating uncommitted changes in a merge", 2017-12-21)
...and again in commit eddd1a411d93 ("merge-recursive: enforce rule that index matches head before merging", 2018-06-30)
...with multiple testcases added to the testsuite that could be
enumerated in even more commits.
Then, finally, in a patch in the same series as the last fix above, the
documentation about this requirement was fixed in commit 55f39cf7551b
("merge: fix misleading pre-merge check documentation", 2018-06-30), and
we all lived happily ever after...
</quick summary>
Unfortunately, "ever after" apparently denotes a limited time and it
expired today. The merge-recursive rule to enforce that index matches
head was at the beginning of merge_trees() and would only trigger when
opt->call_depth was 0. Since merge_recursive() doesn't call
merge_trees() until after returning from recursing, this meant that the
check wasn't triggered by merge_recursive() until it had first finished
all the intermediate merges to create virtual merge bases. That is a
potentially HUGE amount of computation (and writing of intermediate
merge results into the .git/objects directory) before it errors out and
says, in effect, "Sorry, I can't do any merging because you have some
local changes that would be overwritten."
Trying to enforce that all of merge_trees(), merge_recursive(), and
merge_recursive_generic() checked the index == head condition earlier
resulted in a bunch of broken tests. It turns out that
merge_recursive() has code to drop and reload the cache while recursing
to create intermediate virtual merge bases, but unfortunately that code
runs even when no recursion is necessary. This unconditional dropping
and reloading of the cache masked a few bugs:
* builtin/merge-recursive.c: didn't even bother loading the index.
* builtin/stash.c: feels like a fake 'builtin' because it repeatedly
invokes git subprocesses all over the place, mixed with other
operations. In particular, invoking "git reset" will reset the
index on disk, but the parent process that invoked it won't
automatically have its in-memory index updated.
* t3030-merge-recursive.h: this test has always been broken in that it
didn't make sure to make index match head before running. But, it
didn't care about the index or even the merge result, just the
verbose output while running. While commit eddd1a411d93
("merge-recursive: enforce rule that index matches head before
merging", 2018-06-30) should have uncovered this broken test, it
used a test_must_fail wrapper around the merge-recursive call
because it was known that the merge resulted in a rename/rename
conflict. Thus, that fix only made this test fail for a different
reason, and since the index == head check didn't happen until after
coming all the way back out of the recursion, the testcase had
enough information to pass the one check that it did perform.
So, load the index in builtin/merge-recursive.c, reload the in-memory
index in builtin/stash.c, and modify the t3030 testcase to correctly
setup the index and make sure that the test fails in the expected way
(meaning it reports a rename/rename conflict). This makes sure that
all callers actually make the index match head. The next commit will
then enforce the condition that index matches head earlier so this
problem doesn't return in the future.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:27 +0000 (11:41 -0700)]
merge-recursive: remove another implicit dependency on the_repository
Commit d7cf3a96e9a0 ("merge-recursive.c: remove implicit dependency on
the_repository", 2019-01-12) and follow-ups like commit 34e7771bc644
("Use the right 'struct repository' instead of the_repository",
2019-06-27), removed most implicit uses of the_repository. Convert
calls to get_commit_tree() to instead use repo_get_commit_tree() to get
rid of another.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:26 +0000 (11:41 -0700)]
merge-recursive: future-proof update_file_flags() against memory leaks
There is a 'free_buf' label to which all but one of the error paths in
update_file_flags() jump; that error case involves a NULL buf and is
thus not a memory leak. However, make that error case execute the same
deallocation code anyway so that if anyone adds any additional memory
allocations or deallocations, then all error paths correctly deallocate
resources.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Sat, 17 Aug 2019 18:41:24 +0000 (11:41 -0700)]
merge-recursive: provide a better label for diff3 common ancestor
In commit 7ca56aa07619 ("merge-recursive: add a label for ancestor",
2010-03-20), a label was added for the '||||||' line to make it have
the more informative heading '|||||| merged common ancestors', with
the statement:
It would be nicer to use a more informative label. Perhaps someone
will provide one some day.
This chosen label was perfectly reasonable when recursiveness kicks in,
i.e. when there are multiple merge bases. (I can't think of a better
label in such cases.) But it is actually somewhat misleading when there
is a unique merge base or no merge base. Change this based on the
number of merge bases:
>=2: "merged common ancestors"
1: <abbreviated commit hash>
0: "<empty tree>"
Tests have also been added to check that we get the right ancestor name
for each of the three cases.
Also, since merge_recursive() and merge_trees() have polar opposite
pre-conditions for opt->ancestor, document merge_recursive()'s
pre-condition with an assertion. (An assertion was added to
merge_trees() already a few commits ago.) The differences in
pre-conditions stem from two factors: (1) merge_trees() does not recurse
and thus does not have multiple sub-merges to worry about -- each of
which would require a different value for opt->ancestor, (2)
merge_trees() is only passed trees rather than commits and thus cannot
internally guess as good of a label. Thus, while external callers of
merge_trees() are required to provide a non-NULL opt->ancestor,
merge_recursive() expects to set this value itself.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Thu, 15 Aug 2019 21:40:32 +0000 (14:40 -0700)]
merge-recursive: enforce opt->ancestor != NULL when calling merge_trees()
We always want our conflict hunks to be labelled so that users can know
where each came from. The previous commit fixed the one caller in the
codebase which was not setting opt->ancestor (and thus not providing a
label for the "merge base" conflict hunk in diff3-style conflict
markers); add an assertion to prevent future codepaths from also
overlooking this requirement.
Enforcing this requirement also allows us to simplify the code for
labelling the conflict hunks by no longer checking if the ancestor label
is NULL.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Thu, 15 Aug 2019 21:40:31 +0000 (14:40 -0700)]
checkout: provide better conflict hunk description with detached HEAD
When running 'git checkout -m' and using diff3 style conflict markers,
we want all the conflict hunks (left-side, "common" or "merge base", and
right-side) to have label markers letting the user know where each came
from. The "common" hunk label (o.ancestor) came from
old_branch_info->name, but that is NULL when HEAD is detached, which
resulted in a blank label. Check for that case and provide an
abbreviated commit hash instead.
(Incidentally, this was the only case in the git codebase where
merge_trees() was called with opt->ancestor being NULL. A subsequent
commit will prevent similar problems by enforcing that merge_trees()
always be called with opt->ancestor != NULL.)
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren [Thu, 15 Aug 2019 21:40:30 +0000 (14:40 -0700)]
merge-recursive: be consistent with assert
In commit 8daec1df03de ("merge-recursive: switch from (oid,mode) pairs
to a diff_filespec", 2019-04-05), an assertion on a->path && b->path
was added for code readability to document that these both needed to be
non-NULL at this point in the code. However, the subsequent lines also
read o->path, so it should be included in the assert.
Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
projects / rit / log