One of the most common uses of git_path() is to pass a
constant, like git_path("MERGE_MSG"). This has two
drawbacks:
1. The return value is a static buffer, and the lifetime
is dependent on other calls to git_path, etc.
2. There's no compile-time checking of the pathname. This
is OK for a one-off (after all, we have to spell it
correctly at least once), but many of these constant
strings appear throughout the code.
This patch introduces a series of functions to "memoize"
these strings, which are essentially globals for the
lifetime of the program. We compute the value once, take
ownership of the buffer, and return the cached value for
subsequent calls. cache.h provides a helper macro for
defining these functions as one-liners, and defines a few
common ones for global use.
Using a macro is a little bit gross, but it does nicely
document the purpose of the functions. If we need to touch
them all later (e.g., because we learned how to change the
git_dir variable at runtime, and need to invalidate all of
the stored values), it will be much easier to have the
complete list.
Note that the shared-global functions have separate, manual
declarations. We could do something clever with the macros
(e.g., expand it to a declaration in some places, and a
declaration _and_ a definition in path.c). But there aren't
that many, and it's probably better to stay away from
too-magical macros.
Likewise, if we abandon the C preprocessor in favor of
generating these with a script, we could get much fancier.
E.g., normalizing "FOO/BAR-BAZ" into "git_path_foo_bar_baz".
But the small amount of saved typing is probably not worth
the resulting confusion to readers who want to grep for the
function's definition.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Move the interface declaration for the functions in lockfile.c from
cache.h to a new file, lockfile.h. Add #includes where necessary (and
remove some redundant includes of cache.h by files that already
include builtin.h).
Move the documentation of the lock_file state diagram from lockfile.c
to the new header file.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An experiment to use two files (the base file and incremental
changes relative to it) to represent the index to reduce I/O cost
of rewriting a large index when only small part of the working tree
changes.
* nd/split-index: (32 commits)
t1700: new tests for split-index mode
t2104: make sure split index mode is off for the version test
read-cache: force split index mode with GIT_TEST_SPLIT_INDEX
read-tree: note about dropping split-index mode or index version
read-tree: force split-index mode off on --index-output
rev-parse: add --shared-index-path to get shared index path
update-index --split-index: do not split if $GIT_DIR is read only
update-index: new options to enable/disable split index mode
split-index: strip pathname of on-disk replaced entries
split-index: do not invalidate cache-tree at read time
split-index: the reading part
split-index: the writing part
read-cache: mark updated entries for split index
read-cache: save deleted entries in split index
read-cache: mark new entries for split index
read-cache: split-index mode
read-cache: save index SHA-1 after reading
entry.c: update cache_changed if refresh_cache is set in checkout_entry()
cache-tree: mark istate->cache_changed on prime_cache_tree()
cache-tree: mark istate->cache_changed on cache tree update
...
Move "commit->buffer" out of the in-core commit object and keep
track of their lengths. Use this to optimize the code paths to
validate GPG signatures in commit objects.
* jk/commit-buffer-length:
reuse cached commit buffer when parsing signatures
commit: record buffer length in cache
commit: convert commit->buffer to a slab
commit-slab: provide a static initializer
use get_commit_buffer everywhere
convert logmsg_reencode to get_commit_buffer
use get_commit_buffer to avoid duplicate code
use get_cached_commit_buffer where appropriate
provide helpers to access the commit buffer
provide a helper to set the commit buffer
provide a helper to free commit buffer
sequencer: use logmsg_reencode in get_message
logmsg_reencode: return const buffer
do not create "struct commit" with xcalloc
commit: push commit_index update into alloc_commit_node
alloc: include any-object allocations in alloc_report
replace dangerous uses of strbuf_attach
commit_tree: take a pointer/len pair rather than a const strbuf
Like the callsites in the previous commit, logmsg_reencode
already falls back to read_sha1_file when necessary.
However, I split its conversion out into its own commit
because it's a bit more complex.
We return either:
1. The original commit->buffer
2. A newly allocated buffer from read_sha1_file
3. A reencoded buffer (based on either 1 or 2 above).
while trying to do as few extra reads/allocations as
possible. Callers currently free the result with
logmsg_free, but we can simplify this by pointing them
straight to unuse_commit_buffer. This is a slight layering
violation, in that we may be passing a buffer from (3).
However, since the end result is to free() anything except
(1), which is unlikely to change, and because this makes the
interface much simpler, it's a reasonable bending of the
rules.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The return value from logmsg_reencode may be either a newly
allocated buffer or a pointer to the existing commit->buffer.
We would not want the caller to accidentally free() or
modify the latter, so let's mark it as const. We can cast
away the constness in logmsg_free, but only once we have
determined that it is a free-able buffer.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Given that these constants are only being used when updating
references, it is inappropriate to give them such generic names as
"DIE_ON_ERR". So prefix their names with "UPDATE_REFS_".
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git reset" needs to refresh the index when working in a working
tree (it can also be used to match the index to the HEAD in an
otherwise bare repository), but it failed to set up the working
tree properly, causing GIT_WORK_TREE to be ignored.
* nd/reset-setup-worktree:
reset: optionally setup worktree and refresh index on --mixed
Refreshing index requires work tree. So we have two options: always
set up work tree (and refuse to reset if failing to do so), or make
refreshing index optional.
As refreshing index is not the main task, it makes more sense to make
it optional. This allows us to still work in a bare repository to update
what is in the index.
Reported-by: Patrick Palka <patrick@parcs.ath.cx>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When --mixed is used, entries could be removed from index if the
target ref does not have them. When "reset" is used in preparation for
commit spliting (in a dirty worktree), it could be hard to track what
files to be added back. The new option --intent-to-add simplifies it
by marking all removed files intent-to-add.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
The add--interactive --patch mode adjusts the UI based on
whether we are pulling changes from HEAD or elsewhere (in
the former case it asks to unstage the reverse hunk, rather
than apply the forward hunk).
Commit 166ec2e taught reset to work on an unborn branch, but
in doing so, switched to always providing add--interactive
with the sha1 rather than the symbolic name. This meant we
always used the "apply" interface, even for "git reset -p
HEAD".
We can fix this by passing the symbolic name to
add--interactive. Since it understands unborn branches
these days, we do not even have to cover this special case
ourselves; we can simply pass HEAD.
The tests in t7105 now check that the right interface is
used in each circumstance (and notice the regression from
166ec2e we are fixing). The test in t7106 checks that we
get this right for the unborn case, too (not a regression,
since it didn't work at all before, but a nice improvement).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Give "update-refs" a "--stdin" option to read multiple update
requests and perform them in an all-or-none fashion.
* bk/refs-multi-update:
update-ref: add test cases covering --stdin signature
update-ref: support multiple simultaneous updates
refs: add update_refs for multiple simultaneous updates
refs: add function to repack without multiple refs
refs: factor delete_ref loose ref step into a helper
refs: factor update_ref steps into helpers
refs: report ref type from lock_any_ref_for_update
reset: rename update_refs to reset_refs
When using tab-completion, a directory path will often end with a
trailing slash which currently confuses "git reset" when dealing with
submodules. Now that we have parse_pathspec we can easily handle this
by simply adding the PATHSPEC_STRIP_SUBMODULE_SLASH_CHEAP flag.
To do this, we need to move the read_cache() call before the
parse_pathspec() call. All of the existing paths through cmd_reset()
that do not die early already call read_cache() at some point, so there
is no performance impact to doing this in the common case.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git mv A B" when moving a submodule A does "the right thing",
inclusing relocating its working tree and adjusting the paths in
the .gitmodules file.
* jl/submodule-mv: (53 commits)
rm: delete .gitmodules entry of submodules removed from the work tree
mv: update the path entry in .gitmodules for moved submodules
submodule.c: add .gitmodules staging helper functions
mv: move submodules using a gitfile
mv: move submodules together with their work trees
rm: do not set a variable twice without intermediate reading.
t6131 - skip tests if on case-insensitive file system
parse_pathspec: accept :(icase)path syntax
pathspec: support :(glob) syntax
pathspec: make --literal-pathspecs disable pathspec magic
pathspec: support :(literal) syntax for noglob pathspec
kill limit_pathspec_to_literal() as it's only used by parse_pathspec()
parse_pathspec: preserve prefix length via PATHSPEC_PREFIX_ORIGIN
parse_pathspec: make sure the prefix part is wildcard-free
rename field "raw" to "_raw" in struct pathspec
tree-diff: remove the use of pathspec's raw[] in follow-rename codepath
remove match_pathspec() in favor of match_pathspec_depth()
remove init_pathspec() in favor of parse_pathspec()
remove diff_tree_{setup,release}_paths
convert common_prefix() to use struct pathspec
...
After commit 3fde386 (reset [--mixed]: use diff-based reset whether or
not pathspec was given), some code can be moved to the 'reset_type ==
MIXED' check.
Let's move the code that is specific to MIXED.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function resets refs rather than doing arbitrary updates.
Rename it to allow a future general-purpose update_refs function
to be added.
Signed-off-by: Brad King <brad.king@kitware.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This task emerged from b04ba2bb (parse-options: deprecate OPT_BOOLEAN,
2011-09-27). All occurrences of the respective variables have
been reviewed and none of them relied on the counting up mechanism,
but all of them were using the variable as a true boolean.
This patch does not change semantics of any command intentionally.
Signed-off-by: Stefan Beller <stefanbeller@googlemail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This passes the pathspec, more or less unmodified, to
git-add--interactive. The command itself does not process pathspec. It
simply passes the pathspec to other builtin commands. So if all those
commands support pathspec, we're good.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One can set an alias
$ git config [--global] alias.lg "log --graph --pretty=format:'%Cred%h%Creset
-%C(yellow)%d%Creset %s %Cgreen(%cd) %C(bold blue)<%an>%Creset'
--abbrev-commit --date=local"
to see the log as a pretty tree (like *gitk* but in a terminal).
However, log messages written in an encoding i18n.commitEncoding which differs
from terminal encoding are shown corrupted even when i18n.logOutputEncoding
and terminal encoding are the same (e.g. log messages committed on a Cygwin box
with Windows-1251 encoding seen on a Linux box with a UTF-8 encoding and vice versa).
To simplify an example we can say the following two commands are expected
to give the same output to a terminal:
$ git log --oneline --no-color
$ git log --pretty=format:'%h %s'
However, the former pays attention to i18n.logOutputEncoding
configuration, while the latter does not when it formats "%s".
The same corruption is true for
$ git diff --submodule=log
and
$ git rev-list --pretty=format:%s HEAD
and
$ git reset --hard
This patch makes pretty --format honor logOutputEncoding when it formats
log message.
Signed-off-by: Alexey Shumkin <Alex.Crezoff@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When resetting with paths, we no longer require a commit argument, but
only a tree-ish. Update the documentation and synopsis accordingly.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Thanks to b65982b (Optimize "diff-index --cached" using cache-tree,
2009-05-20), resetting with paths is much faster than resetting
without paths. Some timings for the linux-2.6 repo to illustrate this
(best of five, warm cache):
reset reset .
real 0m0.219s 0m0.080s
user 0m0.140s 0m0.040s
sys 0m0.070s 0m0.030s
These two commands should do the same thing, so instead of having the
user type the trailing " ." to get the faster do_diff_cache()-based
implementation, always use it when doing a mixed reset, with or
without paths (so "git reset $rev" would also be faster).
Timing "git reset" shows that it indeed becomes as fast as
"git reset ." after this patch.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some users seem to think, knowingly or not, that being on an unborn
branch is like having a commit with an empty tree checked out, but
when run on an unborn branch, "git reset" currently fails with:
fatal: Failed to resolve 'HEAD' as a valid ref.
Instead of making users figure out that they should run
git rm --cached -r .
, let's teach "git reset" without a revision argument, when on an
unborn branch, to behave as if the user asked to reset to an empty
tree. Don't take the analogy with an empty commit too far, though, but
still disallow explictly referring to HEAD in "git reset HEAD".
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Resetting with paths does not update HEAD and there is nothing else
that a commit should be needed for. Relax the argument parsing so only
a tree is required.
The sha1 is only passed to read_from_tree(), which already only
requires a tree.
The "rev" variable we pass to run_add_interactive() will resolve to a
tree. This is fine since interactive_reset only needs the parameter to
be a treeish and doesn't use it for display purposes.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that there is only one caller left to the single-line method
update_index_refresh(), inline it.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
By not returning from inside the "if (pathspec)" block, we can let the
pathspec-aware and pathspec-less code share a bit more, making it
easier to make future changes that should affect both cases. This also
highlights the similarity between read_from_tree() and reset_index().
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When doing a mixed reset without paths, the index is locked, read,
reset, and written back as part of the actual reset operation (in
reset_index()). Then, when showing the list of worktree modifications,
we lock the index again, refresh it, and write it.
Change this so we only write the index once, making "git reset" a
little faster. It does mean that the index lock will be held a little
longer, but the difference is small compared to the time spent
refreshing the index.
There is one minor functional difference: We used to say "Could not
write new index file." if the first write failed, and "Could not
refresh index" if the second write failed. Now, we will only use the
first message.
This speeds up "git reset" a little on the linux-2.6 repo (best of
five, warm cache):
Before After
real 0m0.239s 0m0.214s
user 0m0.160s 0m0.130s
sys 0m0.070s 0m0.080s
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In preparation for the/a following patch, move the locking, writing
and committing of the index file out of update_index_refresh(). The
code duplication caused will soon be taken care of. What remains of
update_index_refresh() is just one line, but it is still called from
two places, so let's leave it for now.
In the process, we expose and fix the minor UI bug that makes us print
"Could not refresh index" when we fail to write the index file when
invoked with a pathspec. Copy the error message from the pathspec-less
codepath ("Could not write new index file.").
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The final part of cmd_reset() essentially looks like:
if (pathspec) {
...
read_from_tree(...);
} else {
...
reset_index(...);
update_index_refresh(...);
...
}
where read_from_tree() internally also calls
update_index_refresh(). Move the call to update_index_refresh() out of
read_from_tree for symmetry with the 'else' block, making
read_from_tree() and reset_index() closer in functionality.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The switch statement towards the end of reset.c is missing case arms
for KEEP and MERGE for no obvious reason, and soon the only non-empty
case arm will be the one for HARD. So let's proactively replace it by
if-else, which will let us move one if statement out without leaving
funny-looking left-overs.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If writing or committing the new index file fails, we print "Could not
write new index file." followed by "Could not reset index file to
revision $rev.". The first message seems to imply the second, so print
only the first message.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git reset --keep" calls reset_index_file() twice, first doing a
two-way merge to the target revision, updating the index and worktree,
and then resetting the index. After each call, we write the index
file.
In the unlikely event that the second call to reset_index_file()
fails, the index will have been merged to the target revision, but
HEAD will not be updated, leaving the user with a dirty index.
By moving the locking, writing and committing out of
reset_index_file() and into the caller, we can avoid writing the index
twice, thereby making the sure we don't end up in the half-way reset
state. As a bonus, we speed up "git reset --keep" a little on the
linux-2.6 repo (best of five, warm cache):
Before After
real 0m0.315s 0m0.296s
user 0m0.290s 0m0.280s
sys 0m0.020s 0m0.010s
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use a single condition to guard the call to die_if_unmerged_cache for
both --soft and --keep. This avoids the small distraction of the
precondition check from the logic following it.
Also change an instance of
if (e)
err = err || f();
to the almost as short, but clearer
if (e && !err)
err = f();
(which is equivalent since we only care whether exit code is 0)
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
By extracting the code for updating the HEAD and ORIG_HEAD symbolic
references to a separate function, we declutter cmd_reset() a bit and
we make it clear that e.g. the four variables {,sha1_}{,old_}orig are
only used by this code.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Throughout most of parse_args(), the variable 'i' remains at 0. Many
references are still made to the variable even when it could only have
the value 0. This made at least me, who has relatively little
experience with C programming styles, think that parts of the function
was meant to be part of a loop. To avoid such confusion, remove the
variable and also the 'argc' parameter and check for NULL trailing
argv instead.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Declutter cmd_reset() a bit by moving out the argument parsing to its
own function.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Running e.g. "git reset ." in a bare repo results in an index file
being created from the HEAD commit. The differences compared to the
index are then printed as usual, but since there is no worktree, it
will appear as if all files are deleted. For example, in a bare clone
of git.git:
Unstaged changes after reset:
D .gitattributes
D .gitignore
D .mailmap
...
This happens because the check for is_bare_repository() happens after
we branch off into read_from_tree() to reset with paths. Fix by moving
the branching point after the check.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We use the path arguments in two places in reset.c: in
interactive_reset() and read_from_tree(). Both of these call
get_pathspec(), so we pass the (prefix, argv) pair to both
functions. Move the call to get_pathspec() out of these methods, for
two reasons: 1) One argument is simpler than two. 2) It lets us use
the (arguably clearer) "if (pathspec)" in place of "if (i < argc)".
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git reset $pathspec" currently exits with a non-zero exit code if the
worktree is dirty after resetting, which is inconsistent with reset
without pathspec, and it makes it harder to know whether the command
really failed. Change it to exit with code 0 regardless of whether the
worktree is dirty so that non-zero indicates an error.
This makes the 4 "disambiguation" test cases in t7102 clearer since
they all used to "fail", 3 of which "failed" due to changes in the
work tree. Now only the ambiguous one fails.
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 34110cd (Make 'unpack_trees()' have a separate source and
destination index, 2008-03-06), the index no longer gets clobbered by
do_diff_cache() and we can remove the code for discarding and
re-reading it.
There are two paths to update_index_refresh() from cmd_reset(), but on
both paths, either read_cache() or read_cache_unmerged() will have
been called, so the call to read_cache() in this method is redundant
(although practically free).
This speeds up "git reset -- ." a little on the linux-2.6 repo (best
of five, warm cache):
Before After
real 0m0.093s 0m0.080s
user 0m0.040s 0m0.020s
sys 0m0.050s 0m0.050s
Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teaches the object name parser things like a "git describe" output
is always a commit object, "A" in "git log A" must be a committish,
and "A" and "B" in "git log A...B" both must be committish, etc., to
prolong the lifetime of abbreviated object names.
* jc/sha1-name-more: (27 commits)
t1512: match the "other" object names
t1512: ignore whitespaces in wc -l output
rev-parse --disambiguate=<prefix>
rev-parse: A and B in "rev-parse A..B" refer to committish
reset: the command takes committish
commit-tree: the command wants a tree and commits
apply: --build-fake-ancestor expects blobs
sha1_name.c: add support for disambiguating other types
revision.c: the "log" family, except for "show", takes committish
revision.c: allow handle_revision_arg() to take other flags
sha1_name.c: introduce get_sha1_committish()
sha1_name.c: teach lookup context to get_sha1_with_context()
sha1_name.c: many short names can only be committish
sha1_name.c: get_sha1_1() takes lookup flags
sha1_name.c: get_describe_name() by definition groks only commits
sha1_name.c: teach get_short_sha1() a commit-only option
sha1_name.c: allow get_short_sha1() to take other flags
get_sha1(): fix error status regression
sha1_name.c: restructure disambiguation of short names
sha1_name.c: correct misnamed "canonical" and "res"
...
This is not strictly correct, in that resetting selected index
entries from corresponding paths out of a given tree without moving
HEAD is a valid operation, and in such case a tree-ish would suffice.
But the existing code already requires a committish in the codepath,
so let's be consistent with it for now.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
verify_filename() can be called in two different contexts. Either we
just tried to interpret a string as an object name, and it fails, so
we try looking for a working tree file (i.e. we finished looking at
revs that come earlier on the command line, and the next argument
must be a pathname), or we _know_ that we are looking for a
pathname, and shouldn't even try interpreting the string as an
object name.
For example, with this change, we get:
$ git log COPYING HEAD:inexistant
fatal: HEAD:inexistant: no such path in the working tree.
Use '-- <path>...' to specify paths that do not exist locally.
$ git log HEAD:inexistant
fatal: Path 'inexistant' does not exist in 'HEAD'
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
