Make our mergesort implementation type-safe.
* rs/mergesort:
mergesort: remove llist_mergesort()
packfile: use DEFINE_LIST_SORT
fetch-pack: use DEFINE_LIST_SORT
commit: use DEFINE_LIST_SORT
blame: use DEFINE_LIST_SORT
test-mergesort: use DEFINE_LIST_SORT
test-mergesort: use DEFINE_LIST_SORT_DEBUG
mergesort: add macros for typed sort of linked lists
mergesort: tighten merge loop
mergesort: unify ranks loops
Build a typed sort function for blame entries using DEFINE_LIST_SORT
instead of calling llist_mergesort(). This gets rid of the next pointer
accessor functions and their calling overhead at the cost of a slightly
increased object text size.
Before:
__TEXT __DATA __OBJC others dec hex
24621 56 0 147515 172192 2a0a0 blame.o
With this patch:
__TEXT __DATA __OBJC others dec hex
25229 56 0 151702 176987 2b35b blame.o
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce and apply coccinelle rule to discourage an explicit
comparison between a pointer and NULL, and applies the clean-up to
the maintenance track.
* ep/maint-equals-null-cocci:
tree-wide: apply equals-null.cocci
tree-wide: apply equals-null.cocci
contrib/coccinnelle: add equals-null.cocci
Have the diff_free() function call clear_pathspec(). Since the
diff_flush() function calls this all its callers can be simplified to
rely on it instead.
When I added the diff_free() function in e900d494dc (diff: add an API
for deferred freeing, 2021-02-11) I simply missed this, or wasn't
interested in it. Let's consolidate this now. This means that any
future callers (and I've got revision.c in mind) that embed a "struct
diff_options" can simply call diff_free() instead of needing know that
it has an embedded pathspec.
This does fix a bunch of leaks, but I can't mark any test here as
passing under the SANITIZE=leak testing mode because in
886e1084d7 (builtin/: add UNLEAKs, 2017-10-01) an UNLEAK(rev) was
added, which plasters over the memory
leak. E.g. "t4011-diff-symlink.sh" would report fewer leaks with this
fix, but because of the UNLEAK() reports none.
I'll eventually loop around to removing that UNLEAK(rev) annotation as
I'll fix deeper issues with the revisions API leaking. This is one
small step on the way there, a new freeing function in revisions.c
will want to call this diff_free().
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is useful to know when a branch first diverged in history
from some integration branch in order to be able to enumerate
the user's local changes. However, these local changes can
include arbitrary merges, so it is necessary to ignore this
merge structure when finding the divergence point.
In order to do this, teach the "rev-list" family to accept
"--exclude-first-parent-only", which restricts the traversal
of excluded commits to only follow first parent links.
-A-----E-F-G--main
\ / /
B-C-D--topic
In this example, the goal is to return the set {B, C, D} which
represents a topic branch that has been merged into main branch.
`git rev-list topic ^main` will end up returning no commits
since excluding main will end up traversing the commits on topic
as well. `git rev-list --exclude-first-parent-only topic ^main`
however will return {B, C, D} as desired.
Add docs for the new flag, and clarify the doc for --first-parent
to indicate that it applies to traversing the set of included
commits only.
Signed-off-by: Jerry Zhang <jerry@skydio.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Up until recently, object IDs did not have an algorithm member, only a
hash. Consequently, it was possible to share one null (all-zeros)
object ID among all hash algorithms. Now that we're going to be
handling objects from multiple hash algorithms, it's important to make
sure that all object IDs have a correct algorithm field.
Introduce a per-algorithm null OID, and add it to struct hash_algo.
Introduce a wrapper function as well, and use it everywhere we used to
use the null_oid constant.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add and apply a semantic patch for converting code that open-codes
CALLOC_ARRAY to use it instead. It shortens the code and infers the
element size automatically.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A specialization of hashmap that uses a string as key has been
introduced. Hopefully it will see wider use over time.
* en/strmap:
shortlog: use strset from strmap.h
Use new HASHMAP_INIT macro to simplify hashmap initialization
strmap: take advantage of FLEXPTR_ALLOC_STR when relevant
strmap: enable allocations to come from a mem_pool
strmap: add a strset sub-type
strmap: split create_entry() out of strmap_put()
strmap: add functions facilitating use as a string->int map
strmap: enable faster clearing and reusing of strmaps
strmap: add more utility functions
strmap: new utility functions
hashmap: provide deallocation function names
hashmap: introduce a new hashmap_partial_clear()
hashmap: allow re-use after hashmap_free()
hashmap: adjust spacing to fix argument alignment
hashmap: add usage documentation explaining hashmap_free[_entries]()
hashmap_free(), hashmap_free_entries(), and hashmap_free_() have existed
for a while, but aren't necessarily the clearest names, especially with
hashmap_partial_clear() being added to the mix and lazy-initialization
now being supported. Peff suggested we adopt the following names[1]:
- hashmap_clear() - remove all entries and de-allocate any
hashmap-specific data, but be ready for reuse
- hashmap_clear_and_free() - ditto, but free the entries themselves
- hashmap_partial_clear() - remove all entries but don't deallocate
table
- hashmap_partial_clear_and_free() - ditto, but free the entries
This patch provides the new names and converts all existing callers over
to the new naming scheme.
[1] https://lore.kernel.org/git/20201030125059.GA3277724@coredump.intra.peff.net/
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The penultimate commit moved the initialization of 'sb.path' in
'builtin/blame.c::cmd_blame' before the call to
'blame.c::setup_blame_bloom_data'. Since 'cmd_blame' is the only caller
of 'setup_blame_bloom_data', it is now unnecessary for
'setup_blame_bloom_data' to receive 'path' as a separate argument, as
'sb.path' is already initialized.
Remove this argument from setup_blame_bloom_data's interface and use the
'path' field of the 'sb' 'struct blame_scoreboard' instead.
Signed-off-by: Philippe Blain <levraiphilippeblain@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous commit moved the initialization of 'sb.path' in
'builtin/blame.c::cmd_blame' before the call to
'blame.c::setup_scoreboard'. Since 'cmd_blame' is the only caller of
'setup_scoreboard', it is now unnecessary for 'setup_scoreboard' to
receive 'path' as a separate argument, as 'sb.path' is already
initialized.
Remove this argument from setup_scoreboard's interface and use the
'path' field of the 'sb' 'struct blame_scoreboard' instead.
Signed-off-by: Philippe Blain <levraiphilippeblain@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git commit-graph write" learned to limit the number of bloom
filters that are computed from scratch with the --max-new-filters
option.
* tb/bloom-improvements:
commit-graph: introduce 'commitGraph.maxNewFilters'
builtin/commit-graph.c: introduce '--max-new-filters=<n>'
commit-graph: rename 'split_commit_graph_opts'
bloom: encode out-of-bounds filters as non-empty
bloom/diff: properly short-circuit on max_changes
bloom: use provided 'struct bloom_filter_settings'
bloom: split 'get_bloom_filter()' in two
commit-graph.c: store maximum changed paths
commit-graph: respect 'commitGraph.readChangedPaths'
t/helper/test-read-graph.c: prepare repo settings
commit-graph: pass a 'struct repository *' in more places
t4216: use an '&&'-chain
commit-graph: introduce 'get_bloom_filter_settings()'
'get_bloom_filter' takes a flag to control whether it will compute a
Bloom filter if the requested one is missing. In the next patch, we'll
add yet another parameter to this method, which would force all but one
caller to specify an extra 'NULL' parameter at the end.
Instead of doing this, split 'get_bloom_filter' into two functions:
'get_bloom_filter' and 'get_or_compute_bloom_filter'. The former only
looks up a Bloom filter (and does not compute one if it's missing,
thus dropping the 'compute_if_not_present' flag). The latter does
compute missing Bloom filters, with an additional parameter to store
whether or not it needed to do so.
This simplifies many call-sites, since the majority of existing callers
to 'get_bloom_filter' do not want missing Bloom filters to be computed
(so they can drop the parameter entirely and use the simpler version of
the function).
While we're at it, instrument the new 'get_or_compute_bloom_filter()'
with counters in the 'write_commit_graph_context' struct which store
the number of filters that we did and didn't compute, as well as filters
that were truncated.
It would be nice to drop the 'compute_if_not_present' flag entirely,
since all remaining callers of 'get_or_compute_bloom_filter' pass it as
'1', but this will change in a future patch and hence cannot be removed.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many places in the code often need a pointer to the commit-graph's
'struct bloom_filter_settings', in which case they often take the value
from the top-most commit-graph.
In the non-split case, this works as expected. In the split case,
however, things get a little tricky. Not all layers in a chain of
incremental commit-graphs are required to themselves have Bloom data,
and so whether or not some part of the code uses Bloom filters depends
entirely on whether or not the top-most level of the commit-graph chain
has Bloom filters.
This has been the behavior since Bloom filters were introduced, and has
been codified into the tests since a759bfa9ee (t4216: add end to end
tests for git log with Bloom filters, 2020-04-06). In fact, t4216.130
requires that Bloom filters are not used in exactly the case described
earlier.
There is no reason that this needs to be the case, since it is perfectly
valid for commits in an earlier layer to have Bloom filters when commits
in a newer layer do not.
Since Bloom settings are guaranteed in practice to be the same for any
layer in a chain that has Bloom data, it is sufficient to traverse the
'->base_graph' pointer until either (1) a non-null 'struct
bloom_filter_settings *' is found, or (2) until we are at the root of
the commit-graph chain.
Introduce a 'get_bloom_filter_settings()' function that does just this,
and use it instead of purely dereferencing the top-most graph's
'->bloom_filter_settings' pointer.
While we're at it, add an additional test in t5324 to guard against code
in the commit-graph writing machinery that doesn't correctly handle a
NULL 'struct bloom_filter *'.
Co-authored-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
0906ac2b (blame: use changed-path Bloom filters, 2020-04-16)
introduced a call to oidcmp() that should have been oideq(), which
was introduced in 14438c44 (introduce hasheq() and oideq(),
2018-08-28).
Signed-off-by: Edmundo Carmona Antoranz <eantoranz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After blame has finished but before we produce any output, we coalesce
groups of lines that were adjacent in the original suspect (which may
have been split apart by lines in intermediate commits which went away).
However, this can cause incorrect output if the lines are not also
adjacent in the result. For instance, the case in t8003 has:
ABC
DEF
which becomes
ABC
SPLIT
DEF
Blaming only lines 1 and 3 in the result yields two blame groups (one
for each line) that were adjacent in the original. That's enough for us
to coalesce them into a single group, but that loses information: our
output routines assume they're adjacent in the result as well, and we
output:
<oid> 1) ABC
<oid> 2) SPLIT
This is nonsense for two reasons:
- we were asked about line 3, not line 2; we should not output the
SPLIT line at all
- commit <oid> did not touch the SPLIT line at all! We found the
correct blame for line 3, but the bug is actually in the output
stage, which is showing the wrong line number and content from the
final file.
We can fix this by only coalescing when both the suspect and result
lines are adjacent. That fixes this bug, but keeps coalescing in cases
where want it (e.g., the existing test in t8003 where SPLIT goes away,
and the lines really are adjacent in the result).
Reported-by: Nuthan Munaiah <nm6061@rit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We remove members `graph_pos` and `generation` from the struct commit.
The default assignments in init_commit_node() are no longer valid,
which is fine as the slab helpers return appropriate default values and
the assignments are removed.
We will replace existing use of commit->generation and commit->graph_pos
by commit_graph_data_slab helpers using
`contrib/coccinelle/commit.cocci'.
Signed-off-by: Abhishek Kumar <abhishekkumar8222@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We don't use the "parent" parameter at all (probably because the bloom
filter for a commit is always defined against a single parent anyway).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The changed-path Bloom filters help reduce the amount of tree
parsing required during history queries. Before calculating a
diff, we can ask the filter if a path changed between a commit
and its first parent. If the filter says "no" then we can move
on without parsing trees. If the filter says "maybe" then we
parse trees to discover if the answer is actually "yes" or "no".
When computing a blame, there is a section in find_origin() that
computes a diff between a commit and one of its parents. When this
is the first parent, we can check the Bloom filters before calling
diff_tree_oid().
In order to make this work with the blame machinery, we need to
initialize a struct bloom_key with the initial path. But also, we
need to add more keys to a list if a rename is detected. We then
check to see if _any_ of these keys answer "maybe" in the diff.
During development, I purposefully left out this "add a new key
when a rename is detected" to see if the test suite would catch
my error. That is how I discovered the issues with
GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS from the previous change.
With that change, we can feel some confidence in the coverage of
this change.
If a user requests copy detection using "git blame -C", then there
are more places where the set of "important" files can expand. I
do not know enough about how this happens in the blame machinery.
Thus, the Bloom filter integration is explicitly disabled in this
mode. A later change could expand the bloom_key data with an
appropriate call (or calls) to add_bloom_key().
If we did not disable this mode, then the following tests would
fail:
t8003-blame-corner-cases.sh
t8011-blame-split-file.sh
Generally, this is a performance enhancement and should not
change the behavior of 'git blame' in any way. If a repo has a
commit-graph file with computed changed-path Bloom filters, then
they should notice improved performance for their 'git blame'
commands.
Here are some example timings that I found by blaming some paths
in the Linux kernel repository:
git blame arch/x86/kernel/topology.c >/dev/null
Before: 0.83s
After: 0.24s
git blame kernel/time/time.c >/dev/null
Before: 0.72s
After: 0.24s
git blame tools/perf/ui/stdio/hist.c >/dev/null
Before: 0.27s
After: 0.11s
I specifically looked for "deep" paths that were also edited many
times. As a counterpoint, the MAINTAINERS file was edited many
times but is located in the root tree. This means that the cost of
computing a diff relative to the pathspec is very small. Here are
the timings for that command:
git blame MAINTAINERS >/dev/null
Before: 20.1s
After: 18.0s
These timings are the best of five. The worst-case runs were on the
order of 2.5 minutes for both cases. Note that the MAINTAINERS file
has 18,740 lines across 17,000+ commits. This happens to be one of
the cases where this change provides the least improvement.
The lack of improvement for the MAINTAINERS file and the relatively
modest improvement for the other examples can be easily explained.
The blame machinery needs to compute line-level diffs to determine
which lines were changed by each commit. That makes up a large
proportion of the computation time, and this change does not
attempt to improve on that section of the algorithm. The
MAINTAINERS file is large and changed often, so it takes time to
determine which lines were updated by which commit. In contrast,
the code files are much smaller, and it takes longer to comute
the line-by-line diff for a single patch on the Linux mailing
lists.
Outside of the "-C" integration, I believe there is little more to
gain from the changed-path Bloom filters for 'git blame' after this
patch.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since these macros already take a `keyvar' pointer of a known type,
we can rely on OFFSETOF_VAR to get the correct offset without
relying on non-portable `__typeof__' and `offsetof'.
Argument order is also rearranged, so `keyvar' and `member' are
sequential as they are used as: `keyvar->member'
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While we cannot rely on a `__typeof__' operator being portable
to use with `offsetof'; we can calculate the pointer offset
using an existing pointer and the address of a member using
pointer arithmetic for compilers without `__typeof__'.
This allows us to simplify usage of hashmap iterator macros
by not having to specify a type when a pointer of that type
is already given.
In the future, list iterator macros (e.g. list_for_each_entry)
may also be implemented using OFFSETOF_VAR to save hackers the
trouble of using container_of/list_entry macros and without
relying on non-portable `__typeof__'.
v3: use `__typeof__' to avoid clang warnings
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
`hashmap_free_entries' behaves like `container_of' and passes
the offset of the hashmap_entry struct to the internal
`hashmap_free_' function, allowing the function to free any
struct pointer regardless of where the hashmap_entry field
is located.
`hashmap_free' no longer takes any arguments aside from
the hashmap itself.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Inspired by list_for_each_entry in the Linux kernel.
Once again, these are somewhat compromised usability-wise
by compilers lacking __typeof__ support.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update callers to use hashmap_get_entry, hashmap_get_entry_from_hash
or container_of as appropriate.
This is another step towards eliminating the requirement of
hashmap_entry being the first field in a struct.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is less error-prone than "const void *" as the compiler
now detects invalid types being passed.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is less error-prone than "const void *" as the compiler
now detects invalid types being passed.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is less error-prone than "void *" as the compiler now
detects invalid types being passed.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
C compilers do type checking to make life easier for us. So
rely on that and update all hashmap_entry_init callers to take
"struct hashmap_entry *" to avoid future bugs while improving
safety and readability.
Signed-off-by: Eric Wong <e@80x24.org>
Reviewed-by: Derrick Stolee <stolee@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
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>
The tree-walk API learned to pass an in-core repository
instance throughout more codepaths.
* nd/tree-walk-with-repo:
t7814: do not generate same commits in different repos
Use the right 'struct repository' instead of the_repository
match-trees.c: remove the_repo from shift_tree*()
tree-walk.c: remove the_repo from get_tree_entry_follow_symlinks()
tree-walk.c: remove the_repo from get_tree_entry()
tree-walk.c: remove the_repo from fill_tree_descriptor()
sha1-file.c: remove the_repo from read_object_with_reference()
"git blame" learned to "ignore" commits in the history, whose
effects (as well as their presence) get ignored.
* br/blame-ignore:
t8014: remove unnecessary braces
blame: drop some unused function parameters
blame: add a test to cover blame_coalesce()
blame: use the fingerprint heuristic to match ignored lines
blame: add a fingerprint heuristic to match ignored lines
blame: optionally track line fingerprints during fill_blame_origin()
blame: add config options for the output of ignored or unblamable lines
blame: add the ability to ignore commits and their changes
blame: use a helper function in blame_chunk()
Move oidset_parse_file() to oidset.c
fsck: rename and touch up init_skiplist()
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>
This commit integrates the fuzzy fingerprint heuristic into
guess_line_blames().
We actually make two passes. The first pass uses the fuzzy algorithm to
find a match within the current diff chunk. If that fails, the second
pass searches the entire parent file for the best match.
For an example of scanning the entire parent for a match, consider:
commit-a 30) #include <sys/header_a.h>
commit-b 31) #include <header_b.h>
commit-c 32) #include <header_c.h>
Then commit X alphabetizes them:
commit-X 30) #include <header_b.h>
commit-X 31) #include <header_c.h>
commit-X 32) #include <sys/header_a.h>
If we just check the parent's chunk (i.e. the first pass), we'd get:
commit-b 30) #include <header_b.h>
commit-c 31) #include <header_c.h>
commit-X 32) #include <sys/header_a.h>
That's because commit X actually consists of two chunks: one chunk is
removing sys/header_a.h, then some context, and the second chunk is
adding sys/header_a.h.
If we scan the entire parent file, we get:
commit-b 30) #include <header_b.h>
commit-c 31) #include <header_c.h>
commit-a 32) #include <sys/header_a.h>
Signed-off-by: Barret Rhoden <brho@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This algorithm will replace the heuristic used to identify lines from
ignored commits with one that finds likely candidate lines in the
parent's version of the file. The actual replacement occurs in an
upcoming commit.
The old heuristic simply assigned lines in the target to the same line
number (plus offset) in the parent. The new function uses a
fingerprinting algorithm to detect similarity between lines.
The new heuristic is designed to accurately match changes made
mechanically by formatting tools such as clang-format and clang-tidy.
These tools make changes such as breaking up lines to fit within a
character limit or changing identifiers to fit with a naming convention.
The heuristic is not intended to match more extensive refactoring
changes and may give misleading results in such cases.
In most cases formatting tools preserve line ordering, so the heuristic
is optimised for such cases. (Some types of changes do reorder lines
e.g. sorting keep the line content identical, the git blame -M option
can already be used to address this). The reason that it is advantageous
to rely on ordering is due to source code repeating the same character
sequences often e.g. declaring an identifier on one line and using that
identifier on several subsequent lines. This means that lines can look
very similar to each other which presents a problem when doing fuzzy
matching. Relying on ordering gives us extra clues to point towards the
true match.
The heuristic operates on a single diff chunk change at a time. It
creates a “fingerprint” for each line on each side of the change.
Fingerprints are described in detail in the comment for `struct
fingerprint`, but essentially are a multiset of the character pairs in a
line. The heuristic first identifies the line in the target entry whose
fingerprint is most clearly matched to a line fingerprint in the parent
entry. Where fingerprints match identically, the position of the lines
is used as a tie-break. The heuristic locks in the best match, and
subtracts the fingerprint of the line in the target entry from the
fingerprint of the line in the parent entry to prevent other lines being
matched on the same parts of that line. It then repeats the process
recursively on the section of the chunk before the match, and then the
section of the chunk after the match.
Here's an example of the difference the fingerprinting makes. Consider
a file with two commits:
commit-a 1) void func_1(void *x, void *y);
commit-b 2) void func_2(void *x, void *y);
After a commit 'X', we have:
commit-X 1) void func_1(void *x,
commit-X 2) void *y);
commit-X 3) void func_2(void *x,
commit-X 4) void *y);
When we blame-ignored with the old algorithm, we get:
commit-a 1) void func_1(void *x,
commit-b 2) void *y);
commit-X 3) void func_2(void *x,
commit-X 4) void *y);
Where commit-b is blamed for 2 instead of 3. With the fingerprint
algorithm, we get:
commit-a 1) void func_1(void *x,
commit-a 2) void *y);
commit-b 3) void func_2(void *x,
commit-b 4) void *y);
Note line 2 could be matched with either commit-a or commit-b as it is
equally similar to both lines, but is matched with commit-a because its
position as a fraction of the new line range is more similar to commit-a
as a fraction of the old line range. Line 4 is also equally similar to
both lines, but as it appears after line 3 which will be matched first
it cannot be matched with an earlier line.
For many more examples, see t/t8014-blame-ignore-fuzzy.sh which contains
example parent and target files and the line numbers in the parent that
must be matched.
Signed-off-by: Michael Platings <michael@platin.gs>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
fill_blame_origin() is a convenient place to store data that we will use
throughout the lifetime of a blame_origin. Some heuristics for
ignoring commits during a blame session can make use of this storage.
In particular, we will calculate a fingerprint for each line of a file
for blame_origins involved in an ignored commit.
In this commit, we only calculate the line_starts, reusing the existing
code from the scoreboard's line_starts. In an upcoming commit, we will
actually compute the fingerprints.
This feature will be used when we attempt to pass blame entries to
parents when we "ignore" a commit. Most uses of fill_blame_origin()
will not require this feature, hence the flag parameter. Multiple calls
to fill_blame_origin() are idempotent, and any of them can request the
creation of the fingerprints structure.
Suggested-by: Michael Platings <michael@platin.gs>
Signed-off-by: Barret Rhoden <brho@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When ignoring commits, the commit that is blamed might not be
responsible for the change, due to the inaccuracy of our heuristic.
Users might want to know when a particular line has a potentially
inaccurate blame.
Furthermore, guess_line_blames() may fail to find any parent commit for
a given line touched by an ignored commit. Those 'unblamable' lines
remain blamed on an ignored commit. Users might want to know if a line
is unblamable so that they do not spend time investigating a commit they
know is uninteresting.
This patch adds two config options to mark these two types of lines in
the output of blame.
The first option can identify ignored lines by specifying
blame.markIgnoredLines. When this option is set, each blame line that
was blamed on a commit other than the ignored commit is marked with a
'?'.
For example:
278b6158d6fdb (Barret Rhoden 2016-04-11 13:57:54 -0400 26)
appears as:
?278b6158d6fd (Barret Rhoden 2016-04-11 13:57:54 -0400 26)
where the '?' is placed before the commit, and the hash has one fewer
characters.
Sometimes we are unable to even guess at what ancestor commit touched a
line. These lines are 'unblamable.' The second option,
blame.markUnblamableLines, will mark the line with '*'.
For example, say we ignore e5e8d36d04cbe, yet we are unable to blame
this line on another commit:
e5e8d36d04cbe (Barret Rhoden 2016-04-11 13:57:54 -0400 26)
appears as:
*e5e8d36d04cb (Barret Rhoden 2016-04-11 13:57:54 -0400 26)
When these config options are used together, every line touched by an
ignored commit will be marked with either a '?' or a '*'.
Signed-off-by: Barret Rhoden <brho@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commits that make formatting changes or function renames are often not
interesting when blaming a file. A user may deem such a commit as 'not
interesting' and want to ignore and its changes it when assigning blame.
For example, say a file has the following git history / rev-list:
---O---A---X---B---C---D---Y---E---F
Commits X and Y both touch a particular line, and the other commits do
not:
X: "Take a third parameter"
-MyFunc(1, 2);
+MyFunc(1, 2, 3);
Y: "Remove camelcase"
-MyFunc(1, 2, 3);
+my_func(1, 2, 3);
git-blame will blame Y for the change. I'd like to be able to ignore Y:
both the existence of the commit as well as any changes it made. This
differs from -S rev-list, which specifies the list of commits to
process for the blame. We would still process Y, but just don't let the
blame 'stick.'
This patch adds the ability for users to ignore a revision with
--ignore-rev=rev, which may be repeated. They can specify a set of
files of full object names of revs, e.g. SHA-1 hashes, one per line. A
single file may be specified with the blame.ignoreRevFile config option
or with --ignore-rev-file=file. Both the config option and the command
line option may be repeated multiple times. An empty file name "" will
clear the list of revs from previously processed files. Config options
are processed before command line options.
For a typical use case, projects will maintain the file containing
revisions for commits that perform mass reformatting, and their users
have the option to ignore all of the commits in that file.
Additionally, a user can use the --ignore-rev option for one-off
investigation. To go back to the example above, X was a substantive
change to the function, but not the change the user is interested in.
The user inspected X, but wanted to find the previous change to that
line - perhaps a commit that introduced that function call.
To make this work, we can't simply remove all ignored commits from the
rev-list. We need to diff the changes introduced by Y so that we can
ignore them. We let the blames get passed to Y, just like when
processing normally. When Y is the target, we make sure that Y does not
*keep* any blames. Any changes that Y is responsible for get passed to
its parent. Note we make one pass through all of the scapegoats
(parents) to attempt to pass blame normally; we don't know if we *need*
to ignore the commit until we've checked all of the parents.
The blame_entry will get passed up the tree until we find a commit that
has a diff chunk that affects those lines.
One issue is that the ignored commit *did* make some change, and there is
no general solution to finding the line in the parent commit that
corresponds to a given line in the ignored commit. That makes it hard
to attribute a particular line within an ignored commit's diff
correctly.
For example, the parent of an ignored commit has this, say at line 11:
commit-a 11) #include "a.h"
commit-b 12) #include "b.h"
Commit X, which we will ignore, swaps these lines:
commit-X 11) #include "b.h"
commit-X 12) #include "a.h"
We can pass that blame entry to the parent, but line 11 will be
attributed to commit A, even though "include b.h" came from commit B.
The blame mechanism will be looking at the parent's view of the file at
line number 11.
ignore_blame_entry() is set up to allow alternative algorithms for
guessing per-line blames. Any line that is not attributed to the parent
will continue to be blamed on the ignored commit as if that commit was
not ignored. Upcoming patches have the ability to detect these lines
and mark them in the blame output.
The existing algorithm is simple: blame each line on the corresponding
line in the parent's diff chunk. Any lines beyond that stay with the
target.
For example, the parent of an ignored commit has this, say at line 11:
commit-a 11) void new_func_1(void *x, void *y);
commit-b 12) void new_func_2(void *x, void *y);
commit-c 13) some_line_c
commit-d 14) some_line_d
After a commit 'X', we have:
commit-X 11) void new_func_1(void *x,
commit-X 12) void *y);
commit-X 13) void new_func_2(void *x,
commit-X 14) void *y);
commit-c 15) some_line_c
commit-d 16) some_line_d
Commit X nets two additionally lines: 13 and 14. The current
guess_line_blames() algorithm will not attribute these to the parent,
whose diff chunk is only two lines - not four.
When we ignore with the current algorithm, we get:
commit-a 11) void new_func_1(void *x,
commit-b 12) void *y);
commit-X 13) void new_func_2(void *x,
commit-X 14) void *y);
commit-c 15) some_line_c
commit-d 16) some_line_d
Note that line 12 was blamed on B, though B was the commit for
new_func_2(), not new_func_1(). Even when guess_line_blames() finds a
line in the parent, it may still be incorrect.
Signed-off-by: Barret Rhoden <brho@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The same code for splitting a blame_entry at a particular line was used
twice in blame_chunk(), and I'll use the helper again in an upcoming
patch.
Signed-off-by: Barret Rhoden <brho@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git merge-recursive" backend recently learned a new heuristics to
infer file movement based on how other files in the same directory
moved. As this is inherently less robust heuristics than the one
based on the content similarity of the file itself (rather than
based on what its neighbours are doing), it sometimes gives an
outcome unexpected by the end users. This has been toned down to
leave the renamed paths in higher/conflicted stages in the index so
that the user can examine and confirm the result.
* en/merge-directory-renames:
merge-recursive: switch directory rename detection default
merge-recursive: give callers of handle_content_merge() access to contents
merge-recursive: track information associated with directory renames
t6043: fix copied test description to match its purpose
merge-recursive: switch from (oid,mode) pairs to a diff_filespec
merge-recursive: cleanup handle_rename_* function signatures
merge-recursive: track branch where rename occurred in rename struct
merge-recursive: remove ren[12]_other fields from rename_conflict_info
merge-recursive: shrink rename_conflict_info
merge-recursive: move some struct declarations together
merge-recursive: use 'ci' for rename_conflict_info variable name
merge-recursive: rename locals 'o' and 'a' to 'obuf' and 'abuf'
merge-recursive: rename diff_filespec 'one' to 'o'
merge-recursive: rename merge_options argument from 'o' to 'opt'
Use 'unsigned short' for mode, like diff_filespec does
Performance fix around "git blame", especially in a linear history
(which is the norm we should optimize for).
* dk/blame-keep-origin-blob:
blame.c: don't drop origin blobs as eagerly
struct diff_filespec defines mode to be an 'unsigned short'. Several
other places in the API which we'd like to interact with using a
diff_filespec used a plain unsigned (or unsigned int). This caused
problems when taking addresses, so switch to unsigned short.
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a parent blob already has chunks queued up for blaming, dropping
the blob at the end of one blame step will cause it to get reloaded
right away, doubling the amount of I/O and unpacking when processing a
linear history.
Keeping such parent blobs in memory seems like a reasonable optimization
that should incur additional memory pressure mostly when processing the
merges from old branches.
Signed-off-by: David Kastrup <dak@gnu.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Four new configuration variables {author,committer}.{name,email}
have been introduced to override user.{name,email} in more specific
cases.
* wh/author-committer-ident-config:
config: allow giving separate author and committer idents
Elijah Newren
Junio C Hamano
René Scharfe
Ævar Arnfjörð Bjarmason
Jerry Zhang
brian m. carlson
Philippe Blain
Taylor Blau
Edmundo Carmona Antoranz
Jeff King
Abhishek Kumar
Derrick Stolee
Eric Wong
Nguyễn Thái Ngọc Duy
Barret Rhoden
Michael Platings
David Kastrup