A caller of sha1_object_info_extended technically has enough
information to determine the base sha1 from the results of
the call. It knows the pack, offset, and delta type of the
object, which is sufficient to find the base.
However, the functions to do so are not publicly available,
and the code itself is intimate enough with the pack details
that it should be abstracted away. We could add a public
helper to allow callers to query the delta base separately,
but it is simpler and slightly more efficient to optionally
grab it along with the rest of the object_info data.
For cases where the object is not stored as a delta, we
write the null sha1 into the query field. A careful caller
could check "oi.whence == OI_PACKED && oi.u.packed.is_delta"
before looking at the base sha1, but using the null sha1
provides a simple alternative (and gives a better sanity
check for a non-careful caller than simply returning random
bytes).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This parameter is not used yet, but it will be used to tell
sha1_object_info_extended() if it should perform object
replacement or not.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, there is only one caller to lookup_replace_object()
that can benefit from passing it some flags, but we expect
that there could be more.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The READ_SHA1_FILE_REPLACE flag is more related to using the
lookup_replace_object() function rather than the
read_sha1_file() function.
We also need such a flag to be used with sha1_object_info()
instead of read_sha1_file().
The name LOOKUP_REPLACE_OBJECT is therefore better for this
flag.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This may be needed when a hook is run after a new shallow pack is
received, but .git/shallow is not settled yet. A temporary shallow
file to plug all loose ends should be used instead. GIT_SHALLOW_FILE
is overriden by --shallow-file.
--shallow-file does not work in this case because the hook may spawn
many git subprocesses and the launch commands do not have
--shallow-file as it's a recent addition.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Suppose a fetch or push is requested between two shallow repositories
(with no history deepening or shortening). A pack that contains
necessary objects is transferred over together with .git/shallow of
the sender. The receiver has to determine whether it needs to update
.git/shallow if new refs needs new shallow comits.
The rule here is avoid updating .git/shallow by default. But we don't
want to waste the received pack. If the pack contains two refs, one
needs new shallow commits installed in .git/shallow and one does not,
we keep the latter and reject/warn about the former.
Even if .git/shallow update is allowed, we only add shallow commits
strictly necessary for the former ref (remember the sender can send
more shallow commits than necessary) and pay attention not to
accidentally cut the receiver history short (no history shortening is
asked for)
So the steps to figure out what ref need what new shallow commits are:
1. Split the sender shallow commit list into "ours" and "theirs" list
by has_sha1_file. Those that exist in current repo in "ours", the
remaining in "theirs".
2. Check the receiver .git/shallow, remove from "ours" the ones that
also exist in .git/shallow.
3. Fetch the new pack. Either install or unpack it.
4. Do has_sha1_file on "theirs" list again. Drop the ones that fail
has_sha1_file. Obviously the new pack does not need them.
5. If the pack is kept, remove from "ours" the ones that do not exist
in the new pack.
6. Walk the new refs to answer the question "what shallow commits,
both ours and theirs, are required in .git/shallow in order to add
this ref?". Shallow commits not associated to any refs are removed
from their respective list.
7. (*) Check reachability (from the current refs) of all remaining
commits in "ours". Those reachable are removed. We do not want to
cut any part of our (reachable) history. We only check up
commits. True reachability test is done by
check_everything_connected() at the end as usual.
8. Combine the final "ours" and "theirs" and add them all to
.git/shallow. Install new refs. The case where some hook rejects
some refs on a push is explained in more detail in the push
patches.
Of these steps, #6 and #7 are expensive. Both require walking through
some commits, or in the worst case all commits. And we rather avoid
them in at least common case, where the transferred pack does not
contain any shallow commits that the sender advertises. Let's look at
each scenario:
1) the sender has longer history than the receiver
All shallow commits from the sender will be put into "theirs" list
at step 1 because none of them exists in current repo. In the
common case, "theirs" becomes empty at step 4 and exit early.
2) the sender has shorter history than the receiver
All shallow commits from the sender are likely in "ours" list at
step 1. In the common case, if the new pack is kept, we could empty
"ours" and exit early at step 5.
If the pack is not kept, we hit the expensive step 6 then exit
after "ours" is emptied. There'll be only a handful of objects to
walk in fast-forward case. If it's forced update, we may need to
walk to the bottom.
3) the sender has same .git/shallow as the receiver
This is similar to case 2 except that "ours" should be emptied at
step 2 and exit early.
A fetch after "clone --depth=X" is case 1. A fetch after "clone" (from
a shallow repo) is case 3. Luckily they're cheap for the common case.
A push from "clone --depth=X" falls into case 2, which is expensive.
Some more work may be done at the sender/client side to avoid more
work on the server side: if the transferred pack does not contain any
shallow commits, send-pack should not send any shallow commits to the
receive-pack, effectively turning it into a normal push and avoid all
steps.
This patch implements all steps except #3, already handled by
fetch-pack and receive-pack, #6 and #7, which has their own patch due
to their size.
(*) in previous versions step 7 was put before step 3. I reorder it so
that the common case that keeps the pack does not need to walk
commits at all. In future if we implement faster commit
reachability check (maybe with the help of pack bitmaps or commit
cache), step 7 could become cheap and be moved up before 6 again.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new hashmap implementation supports remove, so really remove unused
cache entries from the name hashmap instead of just marking them.
The CE_UNHASHED flag and CE_STATE_MASK are no longer needed.
Keep the CE_HASHED flag to prevent adding entries twice.
Signed-off-by: Karsten Blees <blees@dcon.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Note: the "ce->next = NULL;" in unpack-trees.c::do_add_entry can safely be
removed, as ce->next (now ce->ent.next) is always properly initialized in
name-hash.c::hash_index_entry.
Signed-off-by: Karsten Blees <blees@dcon.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One long-standing flaw in the pack transfer protocol used by "git
clone" was that there was no way to tell the other end which branch
"HEAD" points at, and the receiving end needed to guess. A new
capability has been defined in the pack protocol to convey this
information so that cloning from a repository with more than one
branches pointing at the same commit where the HEAD is at now
reliably sets the initial branch in the resulting repository.
* jc/upload-pack-send-symref:
t5570: Update for clone-progress-to-stderr branch
t5570: Update for symref capability
clone: test the new HEAD detection logic
connect: annotate refs with their symref information in get_remote_head()
connect.c: make parse_feature_value() static
upload-pack: send non-HEAD symbolic refs
upload-pack: send symbolic ref information as capability
upload-pack.c: do not pass confusing cb_data to mark_our_ref()
t5505: fix "set-head --auto with ambiguous HEAD" test
* jx/relative-path-regression-fix:
Use simpler relative_path when set_git_dir
relative_path should honor dos-drive-prefix
test: use unambigous leading path (/foo) for MSYS
"format-patch --from=<whom>" forgot to omit unnecessary in-body from
line, i.e. when <whom> is the same as the real author.
* jk/format-patch-from:
format-patch: print in-body "From" only when needed
This function was added in d2b0708 (2008-09-27, add have_git_dir()
function) as a preparation for adbc0b6 (2008-09-30, cygwin: Use native
Win32 API for stat).
However the second referenced commit was reverted in f66450a (2013-06-22,
cygwin: Remove the Win32 l/stat() implementation), so we don't need to
expose this wrapper function any more as a public API.
Signed-off-by: Stefan Beller <stefanbeller@googlemail.com>
Acked-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `git_open_noatime` helper can be of general interest for other
consumers of git's different on-disk formats.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The said function has this signature:
extern int checkout_entry(struct cache_entry *ce,
const struct checkout *state,
char *topath);
At first glance, it might appear that the caller of checkout_entry()
can specify to which path the contents are written out by the last
parameter, and it is tempting to add "const" in front of its type.
In reality, however, topath[] is to point at a buffer to store the
temporary path generated by the callchain originating from this
function, and the temporary path is always short, much shorter than
the buffer prepared by its only caller in builtin/checkout-index.c.
Document the code a bit to clarify so that future callers know how
to use the function better.
Noticed-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"format-patch --from=<whom>" forgot to omit unnecessary in-body
from line, i.e. when <whom> is the same as the real author.
* jk/format-patch-from:
format-patch: print in-body "From" only when needed
Clean up the internal of the name-hash mechanism used to work
around case insensitivity on some filesystems to cleanly fix a
long-standing API glitch where the caller of cache_name_exists()
that ask about a directory with a counted string was required to
have '/' at one location past the end of the string.
* es/name-hash-no-trailing-slash-in-dirs:
dir: revert work-around for retired dangerous behavior
name-hash: stop storing trailing '/' on paths in index_state.dir_hash
employ new explicit "exists in index?" API
name-hash: refactor polymorphic index_name_exists()
Using a relative_path as git_dir first appears in v1.5.6-1-g044bbbc.
It will make git_dir shorter only if git_dir is inside work_tree,
and this will increase performance. But my last refactor effort on
relative_path function (commit v1.8.3-rc2-12-ge02ca72) changed that.
Always use relative_path as git_dir may bring troubles like
$gmane/234434.
Because new relative_path is a combination of original relative_path
from path.c and original path_relative from quote.c, so in order to
restore the origin implementation, save the original relative_path
as remove_leading_path, and call it in setup.c.
Suggested-by: Karsten Blees <karsten.blees@gmail.com>
Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Instead of typing four capital letters "HEAD", you can say "@" now,
e.g. "git log @".
* fc/at-head:
Add new @ shortcut for HEAD
sha1-name: pass len argument to interpret_branch_name()
Commit a908047 taught format-patch the "--from" option,
which places the author ident into an in-body from header,
and uses the committer ident in the rfc822 from header. The
documentation claims that it will omit the in-body header
when it is the same as the rfc822 header, but the code never
implemented that behavior.
This patch completes the feature by comparing the two idents
and doing nothing when they are the same (this is the same
as simply omitting the in-body header, as the two are by
definition indistinguishable in this case). This makes it
reasonable to turn on "--from" all the time (if it matches
your particular workflow), rather than only using it when
exporting other people's patches.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Depending upon the absence or presence of a trailing '/' on the incoming
pathname, index_name_exists() checks either if a file is present in the
index or if a directory is represented within the index. Each caller
explicitly chooses the mode of operation by adding or removing a
trailing '/' before invoking index_name_exists().
Since these two modes of operations are disjoint and have no code in
common (one searches index_state.name_hash; the other dir_hash), they
can be represented more naturally as distinct functions: one to search
for a file, and one for a directory.
Splitting index searching into two functions relieves callers of the
artificial burden of having to add or remove a slash to select the mode
of operation; instead they just call the desired function. A subsequent
patch will take advantage of this benefit in order to eliminate the
requirement that the incoming pathname for a directory search must have
a trailing slash.
(In order to avoid disturbing in-flight topics, index_name_exists() is
retained as a thin wrapper dispatching either to index_dir_exists() or
index_file_exists().)
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git config" did not provide a way to set or access numbers larger
than a native "int" on the platform; it now provides 64-bit signed
integers on all platforms.
* jk/config-int-range-check:
git-config: always treat --int as 64-bit internally
config: make numeric parsing errors more clear
config: set errno in numeric git_parse_* functions
config: properly range-check integer values
config: factor out integer parsing from range checks
The code that reads from a region that mmaps an on-disk index
assumed that "int"/"short" are always 32/16 bits.
* tg/index-struct-sizes:
read-cache: use fixed width integer types
"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
...
Allow a safer "rewind of the remote tip" push than blind "--force",
by requiring that the overwritten remote ref to be unchanged since
the new history to replace it was prepared.
The machinery is more or less ready. The "--force" option is again
the big red button to override any safety, thanks to J6t's sanity
(the original round allowed --lockref to defeat --force).
The logic to choose the default implemented here is fragile
(e.g. "git fetch" after seeing a failure will update the
remote-tracking branch and will make the next "push" pass,
defeating the safety pretty easily). It is suitable only for the
simplest workflows, and it may hurt users more than it helps them.
* jc/push-cas:
push: teach --force-with-lease to smart-http transport
send-pack: fix parsing of --force-with-lease option
t5540/5541: smart-http does not support "--force-with-lease"
t5533: test "push --force-with-lease"
push --force-with-lease: tie it all together
push --force-with-lease: implement logic to populate old_sha1_expect[]
remote.c: add command line option parser for "--force-with-lease"
builtin/push.c: use OPT_BOOL, not OPT_BOOLEAN
cache.h: move remote/connect API out of it
When you run "git config --int", the maximum size of integer
you get depends on how git was compiled, and what it
considers to be an "int".
This is almost useful, because your scripts calling "git
config" will behave similarly to git internally. But relying
on this is dubious; you have to actually know how git treats
each value internally (e.g., int versus unsigned long),
which is not documented and is subject to change. And even
if you know it is "unsigned long", we do not have a
git-config option to match that behavior.
Furthermore, you may simply be asking git to store a value
on your behalf (e.g., configuration for a hook). In that
case, the relevant range check has nothing at all to do with
git, but rather with whatever scripting tools you are using
(and git has no way of knowing what the appropriate range is
there).
Not only is the range check useless, but it is actively
harmful, as there is no way at all for scripts to look
at config variables with large values. For instance, one
cannot reliably get the value of pack.packSizeLimit via
git-config. On an LP64 system, git happily uses a 64-bit
"unsigned long" internally to represent the value, but the
script cannot read any value over 2G.
Ideally, the "--int" option would simply represent an
arbitrarily large integer. For practical purposes, however,
a 64-bit integer is large enough, and is much easier to
implement (and if somebody overflows it, we will still
notice the problem, and not simply return garbage).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is useful to make sure we don't step outside the boundaries of what
we are interpreting at the moment. For example while interpreting
foobar@{u}~1, the job of interpret_branch_name() ends right before ~1,
but there's no way to figure that out inside the function, unless the
len argument is passed.
So let's do that.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use the fixed width integer types uint16_t and uint32_t for on-disk
structures; unsigned short and unsigned int do not have a guaranteed
size.
Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If somebody wants to only know on-disk footprint of an object
without having to know its type or payload size, we can bypass a
lot of code to cheaply learn it.
* jk/cat-file-batch-optim:
Fix some sparse warnings
sha1_object_info_extended: pass object_info to helpers
sha1_object_info_extended: make type calculation optional
packed_object_info: make type lookup optional
packed_object_info: hoist delta type resolution to helper
sha1_loose_object_info: make type lookup optional
sha1_object_info_extended: rename "status" to "type"
cat-file: disable object/refname ambiguity check for batch mode
Add "interactive" mode to "git clean".
The early part to refactor relative path related helper functions
looked sensible.
* jx/clean-interactive:
test: run testcases with POSIX absolute paths on Windows
test: add t7301 for git-clean--interactive
git-clean: add documentation for interactive git-clean
git-clean: add ask each interactive action
git-clean: add select by numbers interactive action
git-clean: add filter by pattern interactive action
git-clean: use a git-add-interactive compatible UI
git-clean: add colors to interactive git-clean
git-clean: show items of del_list in columns
git-clean: add support for -i/--interactive
git-clean: refactor git-clean into two phases
write_name{_quoted_relative,}(): remove redundant parameters
quote_path_relative(): remove redundant parameter
quote.c: substitute path_relative with relative_path
path.c: refactor relative_path(), not only strip prefix
test: add test cases for relative_path
Allow configuration data to be read from in-tree blob objects,
which would help working in a bare repository and submodule
updates.
* hv/config-from-blob:
do not die when error in config parsing of buf occurs
teach config --blob option to parse config from database
config: make parsing stack struct independent from actual data source
config: drop cf validity check in get_next_char()
config: factor out config file stack management
When "git" is spawned in such a way that any of the low 3 file
descriptors is closed, our first open() may yield file descriptor 2,
and writing error message to it would screw things up in a big way.
* tr/protect-low-3-fds:
git: ensure 0/1/2 are open in main()
daemon/shell: refactor redirection of 0/1/2 from /dev/null
"git cat-file --batch-check=<format>" is added, primarily to allow
on-disk footprint of objects in packfiles (often they are a lot
smaller than their true size, when expressed as deltas) to be
reported.
* jk/in-pack-size-measurement:
pack-revindex: radix-sort the revindex
pack-revindex: use unsigned to store number of objects
cat-file: split --batch input lines on whitespace
cat-file: add %(objectsize:disk) format atom
cat-file: add --batch-check=<format>
cat-file: refactor --batch option parsing
cat-file: teach --batch to stream blob objects
t1006: modernize output comparisons
teach sha1_object_info_extended a "disk_size" query
zero-initialize object_info structs
Both daemon.c and shell.c contain logic to open FDs 0/1/2 from
/dev/null if they are not already open. Move the function in daemon.c
to setup.c and use it in shell.c, too.
While there, remove a 'not' that inverted the meaning of the comment.
The point is indeed to *avoid* messing up.
Signed-off-by: Thomas Rast <trast@inf.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
:(glob)path differs from plain pathspec that it uses wildmatch with
WM_PATHNAME while the other uses fnmatch without FNM_PATHNAME. The
difference lies in how '*' (and '**') is processed.
With the introduction of :(glob) and :(literal) and their global
options --[no]glob-pathspecs, the user can:
- make everything literal by default via --noglob-pathspecs
--literal-pathspecs cannot be used for this purpose as it
disables _all_ pathspec magic.
- individually turn on globbing with :(glob)
- make everything globbing by default via --glob-pathspecs
- individually turn off globbing with :(literal)
The implication behind this is, there is no way to gain the default
matching behavior (i.e. fnmatch without FNM_PATHNAME). You either get
new globbing or literal. The old fnmatch behavior is considered
deprecated and discouraged to use.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prepending prefix to pathspec is a trick to workaround the fact that
commands can be executed in a subdirectory, but all git commands run
at worktree's root. The prefix part should always be treated as
literal string. Make it so.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Each caller of sha1_object_info_extended sets up an
object_info struct to tell the function which elements of
the object it wants to get. Until now, getting the type of
the object has always been required (and it is returned via
the return type rather than a pointer in object_info).
This can involve actually opening a loose object file to
determine its type, or following delta chains to determine a
packed file's base type. These effects produce a measurable
slow-down when doing a "cat-file --batch-check" that does
not include %(objecttype).
This patch adds a "typep" query to struct object_info, so
that it can be optionally queried just like size and
disk_size. As a result, the return type of the function is
no longer the object type, but rather 0/-1 for success/error.
As there are only three callers total, we just fix up each
caller rather than keep a compatibility wrapper:
1. The simpler sha1_object_info wrapper continues to
always ask for and return the type field.
2. The istream_source function wants to know the type, and
so always asks for it.
3. The cat-file batch code asks for the type only when
%(objecttype) is part of the format string.
On linux.git, the best-of-five for running:
$ git rev-list --objects --all >objects
$ time git cat-file --batch-check='%(objectsize:disk)'
on a fully packed repository goes from:
real 0m8.680s
user 0m8.160s
sys 0m0.512s
to:
real 0m7.205s
user 0m6.580s
sys 0m0.608s
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A common use of "cat-file --batch-check" is to feed a list
of objects from "rev-list --objects" or a similar command.
In this instance, all of our input objects are 40-byte sha1
ids. However, cat-file has always allowed arbitrary revision
specifiers, and feeds the result to get_sha1().
Fortunately, get_sha1() recognizes a 40-byte sha1 before
doing any hard work trying to look up refs, meaning this
scenario should end up spending very little time converting
the input into an object sha1. However, since 798c35f
(get_sha1: warn about full or short object names that look
like refs, 2013-05-29), when we encounter this case, we
spend the extra effort to do a refname lookup anyway, just
to print a warning. This is further exacerbated by ca91993
(get_packed_ref_cache: reload packed-refs file when it
changes, 2013-06-20), which makes individual ref lookup more
expensive by requiring a stat() of the packed-refs file for
each missing ref.
With no patches, this is the time it takes to run:
$ git rev-list --objects --all >objects
$ time git cat-file --batch-check='%(objectname)' <objects
on the linux.git repository:
real 1m13.494s
user 0m25.924s
sys 0m47.532s
If we revert ca91993, the packed-refs up-to-date check, it
gets a little better:
real 0m54.697s
user 0m21.692s
sys 0m32.916s
but we are still spending quite a bit of time on ref lookup
(and we would not want to revert that patch, anyway, which
has correctness issues). If we revert 798c35f, disabling
the warning entirely, we get a much more reasonable time:
real 0m7.452s
user 0m6.836s
sys 0m0.608s
This patch does the moral equivalent of this final case (and
gets similar speedups). We introduce a global flag that
callers of get_sha1() can use to avoid paying the price for
the warning.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This can be used to read configuration values directly from git's
database. For example it is useful for reading to be checked out
.gitmodules files directly from the database.
Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I attempted to make index_state->cache[] a "const struct cache_entry **"
to find out how existing entries in index are modified and where. The
question I have is what do we do if we really need to keep track of on-disk
changes in the index. The result is
- diff-lib.c: setting CE_UPTODATE
- name-hash.c: setting CE_HASHED
- preload-index.c, read-cache.c, unpack-trees.c and
builtin/update-index: obvious
- entry.c: write_entry() may refresh the checked out entry via
fill_stat_cache_info(). This causes "non-const struct cache_entry
*" in builtin/apply.c, builtin/checkout-index.c and
builtin/checkout.c
- builtin/ls-files.c: --with-tree changes stagemask and may set
CE_UPDATE
Of these, write_entry() and its call sites are probably most
interesting because it modifies on-disk info. But this is stat info
and can be retrieved via refresh, at least for porcelain
commands. Other just uses ce_flags for local purposes.
So, keeping track of "dirty" entries is just a matter of setting a
flag in index modification functions exposed by read-cache.c. Except
unpack-trees, the rest of the code base does not do anything funny
behind read-cache's back.
The actual patch is less valueable than the summary above. But if
anyone wants to re-identify the above sites. Applying this patch, then
this:
diff --git a/cache.h b/cache.h
index 430d021..1692891 100644
--- a/cache.h
+++ b/cache.h
@@ -267,7 +267,7 @@ static inline unsigned int canon_mode(unsigned int mode)
#define cache_entry_size(len) (offsetof(struct cache_entry,name) + (len) + 1)
struct index_state {
- struct cache_entry **cache;
+ const struct cache_entry **cache;
unsigned int version;
unsigned int cache_nr, cache_alloc, cache_changed;
struct string_list *resolve_undo;
will help quickly identify them without bogus warnings.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The definition of "struct ref" in "cache.h", a header file so
central to the system, always confused me. This structure is not
about the local ref used by sha1-name API to name local objects.
It is what refspecs are expanded into, after finding out what refs
the other side has, to define what refs are updated after object
transfer succeeds to what values. It belongs to "remote.h" together
with "struct refspec".
While we are at it, also move the types and functions related to the
Git transport connection to a new header file connect.h
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Using sha1_object_info_extended, a caller can find out the
type of an object, its size, and information about where it
is stored. In addition to the object's "true" size, it can
also be useful to know the size that the object takes on
disk (e.g., to generate statistics about which refs consume
space).
This patch adds a "disk_sizep" field to "struct object_info",
and fills it in during sha1_object_info_extended if it is
non-NULL.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git pack-refs" that races with new ref creation or deletion have
been susceptible to lossage of refs under right conditions, which
has been tightened up.
* mh/ref-races:
for_each_ref: load all loose refs before packed refs
get_packed_ref_cache: reload packed-refs file when it changes
add a stat_validity struct
Extract a struct stat_data from cache_entry
packed_ref_cache: increment refcount when locked
do_for_each_entry(): increment the packed refs cache refcount
refs: manage lifetime of packed refs cache via reference counting
refs: implement simple transactions for the packed-refs file
refs: wrap the packed refs cache in a level of indirection
pack_refs(): split creation of packed refs and entry writing
repack_without_ref(): split list curation and entry writing
Original design of relative_path() is simple, just strip the prefix
(*base) from the absolute path (*abs).
In most cases, we need a real relative path, such as: ../foo,
../../bar. That's why there is another reimplementation
(path_relative()) in quote.c.
Borrow some codes from path_relative() in quote.c to refactor
relative_path() in path.c, so that it could return real relative
path, and user can reuse this function without reimplementing
his/her own. The function path_relative() in quote.c will be
substituted, and I would use the new relative_path() function when
implementing the interactive git-clean later.
Different results for relative_path() before and after this refactor:
abs path base path relative (original) relative (refactor)
======== ========= =================== ===================
/a/b /a/b . ./
/a/b/ /a/b . ./
/a /a/b/ /a ../
/ /a/b/ / ../../
/a/c /a/b/ /a/c ../c
/x/y /a/b/ /x/y ../../x/y
a/b/ a/b/ . ./
a/b/ a/b . ./
a a/b a ../
x/y a/b/ x/y ../../x/y
a/c a/b a/c ../c
(empty) (null) (empty) ./
(empty) (empty) (empty) ./
(empty) /a/b (empty) ./
(null) (null) (null) ./
(null) (empty) (null) ./
(null) /a/b (segfault) ./
You may notice that return value "." has been changed to "./".
It is because:
* Function quote_path_relative() in quote.c will show the relative
path as "./" if abs(in) and base(prefix) are the same.
* Function relative_path() is called only once (in setup.c), and
it will be OK for the return value as "./" instead of ".".
Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It can sometimes be useful to know whether a path in the
filesystem has been updated without going to the work of
opening and re-reading its content. We trust the stat()
information on disk already to handle index updates, and we
can use the same trick here.
This patch introduces a "stat_validity" struct which
encapsulates the concept of checking the stat-freshness of a
file. It is implemented on top of "struct stat_data" to
reuse the logic about which stat entries to trust for a
particular platform, but hides the complexity behind two
simple functions: check and update.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add public functions fill_stat_data() and match_stat_data() to work
with it. This infrastructure will later be used to check the validity
of other types of file.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* rs/unpack-trees-plug-leak:
unpack-trees: free cache_entry array members for merges
diff-lib, read-tree, unpack-trees: mark cache_entry array paramters const
diff-lib, read-tree, unpack-trees: mark cache_entry pointers const
unpack-trees: create working copy of merge entry in merged_entry
unpack-trees: factor out dup_entry
read-cache: mark cache_entry pointers const
cache: mark cache_entry pointers const
5f44324 (core: log offset pack data accesses happened - 2011-07-06)
provides a way to observe pack access patterns via a config
switch. Setting an environment variable looks more obvious than a
config var, especially when you just need to _observe_, and more
inline with other tracing knobs we have.
Document it as it may be useful for remote troubleshooting.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git fetch origin master" unlike "git fetch origin" or "git fetch"
did not update "refs/remotes/origin/master"; this was an early
design decision to keep the update of remote tracking branches
predictable, but in practice it turns out that people find it more
convenient to opportunisticly update them whenever we have a chance,
and we have been updating them when we run "git push" which already
breaks the original "predictability" anyway.
Now such a fetch does update refs/remotes/origin/master.
* jk/fetch-always-update-tracking:
fetch: don't try to update unfetched tracking refs
fetch: opportunistically update tracking refs
refactor "ref->merge" flag
fetch/pull doc: untangle meaning of bare <ref>
t5510: start tracking-ref tests from a known state
ie_match_stat and ie_modified only derefence their struct cache_entry
pointers for reading. Add const to the parameter declaration here and
do the same for the static helper function used by them, as it's the
same there as well. This allows callers to pass in const pointers.
Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add const for pointers that are only dereferenced for reading by the
inline functions copy_cache_entry and ce_mode_from_stat. This allows
callers to pass in const pointers.
Signed-off-by: René Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We used the approxidate() parser for "--expire=<timestamp>" options
of various commands, but it is better to treat --expire=all and
--expire=now a bit more specially than using the current timestamp.
Update "git gc" and "git reflog" with a new parsing function for
expiry dates.
* jc/prune-all:
prune: introduce OPT_EXPIRY_DATE() and use it
api-parse-options.txt: document "no-" for non-boolean options
git-gc.txt, git-reflog.txt: document new expiry options
date.c: add parse_expiry_date()
Each "struct ref" has a boolean flag that is set by the
fetch code to determine whether the ref should be marked as
"not-for-merge" or not when we write it out to FETCH_HEAD.
It would be useful to turn this boolean into a tri-state,
with the third state meaning "do not bother writing it out
to FETCH_HEAD at all". That would let us add extra refs to
the set of refs to be stored (e.g., to store copies of
things we fetched) without impacting FETCH_HEAD.
This patch turns it into an enum that covers the tri-state
case, and hopefully makes the code more explicit and easier
to read.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As promised in 0fa2eb530f (add: warn when -u or -A is used without
pathspec, 2013-01-28), in Git 2.0, "git add -u/-A" that is run
without pathspec in a subdirectory updates all updated paths in the
entire working tree, not just the current directory and its
subdirectories.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reduce duplicated code between convert.c and attr.c.
* lf/read-blob-data-from-index:
convert.c: remove duplicate code
read_blob_data_from_index(): optionally return the size of blob data
attr.c: extract read_index_data() as read_blob_data_from_index()
In Git 2.0, "git add -u" and "git add -A" without any pathspec will
update the index for all paths, including those outside the current
directory, making it more consistent with "commit -a". To help the
migration pain, a warning is issued when the differences between the
current behaviour and the upcoming behaviour matters, i.e. when the
user has local changes outside the current directory.
* 'jn/add-2.0-u-A-sans-pathspec' (early part):
add -A: only show pathless 'add -A' warning when changes exist outside cwd
add -u: only show pathless 'add -u' warning when changes exist outside cwd
add: make warn_pathless_add() a no-op after first call
add: add a blank line at the end of pathless 'add [-u|-A]' warning
add: make pathless 'add [-u|-A]' warning a file-global function
"git reflog --expire=all" tries to expire reflog entries up to the
current second, because the approxidate() parser gives the current
timestamp for anything it does not understand (and it does not know
what time "all" means). When the user tells us to expire "all" (or
set the expiration time to "now"), the user wants to remove all the
reflog entries (no reflog entry should record future time).
Just set it to ULONG_MAX and to let everything that is older that
timestamp expire.
While at it, allow "now" to be treated the same way for callers that
parse expiry date timestamp with this function. Also use an error
reporting version of approxidate() to report misspelled date. When
the user says e.g. "--expire=mnoday" to delete entries two days or
older on Wednesday, we wouldn't want the "unknown, default to now"
logic to kick in.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This allows for optionally getting the size of the returned data and
will be used in a follow-up patch.
Signed-off-by: Lukas Fleischer <git@cryptocrack.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extract the read_index_data() function from attr.c and move it to
read-cache.c; rename it to read_blob_data_from_index() and update
the function signature of it to align better with index/cache API
functions.
This allows for reusing the function in convert.c later.
Signed-off-by: Lukas Fleischer <git@cryptocrack.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
All calls to set_shared_perm() use mode == 0, so simplify the
function.
Because all callers use the macro adjust_shared_perm(path) from
cache.h to call this function, convert it to a proper function,
losing set_shared_perm().
Since path.c has much more functions than just mkpath() these days,
drop the stale comment about it.
Signed-off-by: Torsten Bögershausen <tboegi@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A common workflow in large projects is to chdir into a subdirectory of
interest and only do work there:
cd src
vi foo.c
make test
git add -u
git commit
The upcoming change to 'git add -u' behavior would not affect such a
workflow: when the only changes present are in the current directory,
'git add -u' will add all changes, and whether that happens via an
implicit "." or implicit ":/" parameter is an unimportant
implementation detail.
The warning about use of 'git add -u' with no pathspec is annoying
because it seemingly serves no purpose in this case. So suppress the
warning unless there are changes outside the cwd that are not being
added.
A previous version of this patch ran two I/O-intensive diff-files
passes: one to find changes outside the cwd, and another to find
changes to add to the index within the cwd. This version runs one
full-tree diff and decides for each change whether to add it or warn
and suppress it in update_callback. As a result, even on very large
repositories "git add -u" will not be significantly slower than the
future default behavior ("git add -u :/"), and the slowdown relative
to "git add -u ." should be a useful clue to users of such
repositories to get into the habit of explicitly passing '.'.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Acked-by: Jeff King <peff@peff.net>
Improved-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Consolidate repeated pathspec matches on the same paths, while
fixing a bug in "git checkout dir/" code started from an unmerged
index.
* nd/checkout-paths-reduce-match-pathspec-calls:
checkout: avoid unnecessary match_pathspec calls
An aliased command spawned from a bare repository that does not say
it is bare with "core.bare = yes" is treated as non-bare by mistake.
* jk/alias-in-bare:
setup: suppress implicit "." work-tree for bare repos
environment: add GIT_PREFIX to local_repo_env
cache.h: drop LOCAL_REPO_ENV_SIZE
* maint-1.8.1:
Start preparing for 1.8.1.6
git-tag(1): we tag HEAD by default
Fix revision walk for commits with the same dates
t2003: work around path mangling issue on Windows
pack-refs: add fully-peeled trait
pack-refs: write peeled entry for non-tags
use parse_object_or_die instead of die("bad object")
avoid segfaults on parse_object failure
entry: fix filter lookup
t2003: modernize style
name-hash.c: fix endless loop with core.ignorecase=true
The code to keep track of what directory names are known to Git on
platforms with case insensitive filesystems can get confused upon
a hash collision between these pathnames and looped forever.
* kb/name-hash:
name-hash.c: fix endless loop with core.ignorecase=true
Clean up pkt-line API, implementation and its callers to make them
more robust.
* jk/pkt-line-cleanup:
do not use GIT_TRACE_PACKET=3 in tests
remote-curl: always parse incoming refs
remote-curl: move ref-parsing code up in file
remote-curl: pass buffer straight to get_remote_heads
teach get_remote_heads to read from a memory buffer
pkt-line: share buffer/descriptor reading implementation
pkt-line: provide a LARGE_PACKET_MAX static buffer
pkt-line: move LARGE_PACKET_MAX definition from sideband
pkt-line: teach packet_read_line to chomp newlines
pkt-line: provide a generic reading function with options
pkt-line: drop safe_write function
pkt-line: move a misplaced comment
write_or_die: raise SIGPIPE when we get EPIPE
upload-archive: use argv_array to store client arguments
upload-archive: do not copy repo name
send-pack: prefer prefixcmp over memcmp in receive_status
fetch-pack: fix out-of-bounds buffer offset in get_ack
upload-pack: remove packet debugging harness
upload-pack: do not add duplicate objects to shallow list
upload-pack: use get_sha1_hex to parse "shallow" lines
In checkout_paths() we do this
- for all updated items, call match_pathspec
- for all items, call match_pathspec (inside unmerge_cache)
- for all items, call match_pathspec (for showing "path .. is unmerged)
- for updated items, call match_pathspec and update paths
That's a lot of duplicate match_pathspec(s) and the function is not
exactly cheap to be called so many times, especially on large indexes.
This patch makes it call match_pathspec once per updated index entry,
save the result in ce_flags and reuse the results in the following
loops.
The changes in 0a1283b (checkout $tree $path: do not clobber local
changes in $path not in $tree - 2011-09-30) limit the affected paths
to ones we read from $tree. We do not do anything to other modified
entries in this case, so the "for all items" above could be modified
to "for all updated items". But..
The command's behavior now is modified slightly: unmerged entries that
match $path, but not updated by $tree, are now NOT touched. Although
this should be considered a bug fix, not a regression. A new test is
added for this change.
And while at there, free ps_matched after use.
The following command is tested on webkit, 215k entries. The pattern
is chosen mainly to make match_pathspec sweat:
git checkout -- "*[a-zA-Z]*[a-zA-Z]*[a-zA-Z]*"
before after
real 0m3.493s 0m2.737s
user 0m2.239s 0m1.586s
sys 0m1.252s 0m1.151s
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An aliased command spawned from a bare repository that does not say
it is bare with "core.bare = yes" is treated as non-bare by mistake.
* jk/alias-in-bare:
setup: suppress implicit "." work-tree for bare repos
environment: add GIT_PREFIX to local_repo_env
cache.h: drop LOCAL_REPO_ENV_SIZE
"git count-objects -v" did not count leftover temporary packfiles
and other kinds of garbage.
* nd/count-garbage:
count-objects: report how much disk space taken by garbage files
count-objects: report garbage files in pack directory too
sha1_file: reorder code in prepare_packed_git_one()
git-count-objects.txt: describe each line in -v output
Allows requests to fetch objects at any tip of refs (including
hidden ones). It seems that there may be use cases even outside
Gerrit (e.g. $gmane/215701).
* jc/fetch-raw-sha1:
fetch: fetch objects by their exact SHA-1 object names
upload-pack: optionally allow fetching from the tips of hidden refs
fetch: use struct ref to represent refs to be fetched
parse_fetch_refspec(): clarify the codeflow a bit
We use the function git_deflate_init() -- which wraps the zlib function
deflateInit() -- to initialize compression of ZIP file entries. This
results in compressed data prefixed with a two-bytes long header and
followed by a four-bytes trailer. ZIP file entries consist of ZIP
headers and raw compressed data instead, so we remove the zlib wrapper
before writing the result.
We can ask zlib for the the raw compressed data without the unwanted
parts in the first place by using deflateInit2() and specifying a
negative number of bits to size the window. For that purpose, factor
out the function do_git_deflate_init() and add git_deflate_init_raw(),
which wraps it. Then use the latter in archive-zip.c and get rid of
the code that stripped the zlib header and trailer.
Also rename the helper function zlib_deflate() to zlib_deflate_raw()
to reflect the change.
Thus we avoid generating data that we throw away anyway, the code
becomes shorter and some magic constants are removed.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If an explicit GIT_DIR is given without a working tree, we
implicitly assume that the current working directory should
be used as the working tree. E.g.,:
GIT_DIR=/some/repo.git git status
would compare against the cwd.
Unfortunately, we fool this rule for sub-invocations of git
by setting GIT_DIR internally ourselves. For example:
git init foo
cd foo/.git
git status ;# fails, as we expect
git config alias.st status
git status ;# does not fail, but should
What happens is that we run setup_git_directory when doing
alias lookup (since we need to see the config), set GIT_DIR
as a result, and then leave GIT_WORK_TREE blank (because we
do not have one). Then when we actually run the status
command, we do setup_git_directory again, which sees our
explicit GIT_DIR and uses the cwd as an implicit worktree.
It's tempting to argue that we should be suppressing that
second invocation of setup_git_directory, as it could use
the values we already found in memory. However, the problem
still exists for sub-processes (e.g., if "git status" were
an external command).
You can see another example with the "--bare" option, which
sets GIT_DIR explicitly. For example:
git init foo
cd foo/.git
git status ;# fails
git --bare status ;# does NOT fail
We need some way of telling sub-processes "even though
GIT_DIR is set, do not use cwd as an implicit working tree".
We could do it by putting a special token into
GIT_WORK_TREE, but the obvious choice (an empty string) has
some portability problems.
Instead, we add a new boolean variable, GIT_IMPLICIT_WORK_TREE,
which suppresses the use of cwd as a working tree when
GIT_DIR is set. We trigger the new variable when we know we
are in a bare setting.
The variable is left intentionally undocumented, as this is
an internal detail (for now, anyway). If somebody comes up
with a good alternate use for it, and once we are confident
we have shaken any bugs out of it, we can consider promoting
it further.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The GIT_PREFIX variable is set based on our location within
the working tree. It should therefore be cleared whenever
GIT_WORK_TREE is cleared.
In practice, this doesn't cause any bugs, because none of
the sub-programs we invoke with local_repo_env cleared
actually care about GIT_PREFIX. But this is the right thing
to do, and future proofs us against that assumption changing.
While we're at it, let's define a GIT_PREFIX_ENVIRONMENT
macro; this avoids repetition of the string literal, which
can help catch any spelling mistakes in the code.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We keep a static array of variables that should be cleared
when invoking a sub-process on another repo. We statically
size the array with the LOCAL_REPO_ENV_SIZE macro so that
any readers do not have to count it themselves.
As it turns out, no readers actually use the macro, and it
creates a maintenance headache, as modifications to the
array need to happen in two places (one to add the new
element, and another to bump the size).
Since it's NULL-terminated, we can just drop the size macro
entirely. While we're at it, we'll clean up some comments
around it, and add a new mention of it at the top of the
list of environment variable macros. Even though
local_repo_env is right below that list, it's easy to miss,
and additions to that list should consider local_repo_env.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With core.ignorecase=true, name-hash.c builds a case insensitive index of
all tracked directories. Currently, the existing cache entry structures are
added multiple times to the same hashtable (with different name lengths and
hash codes). However, there's only one dir_next pointer, which gets
completely messed up in case of hash collisions. In the worst case, this
causes an endless loop if ce == ce->dir_next (see t7062).
Use a separate hashtable and separate structures for the directory index
so that each directory entry has its own next pointer. Use reference
counting to track which directory entry contains files.
There are only slight changes to the name-hash.c API:
- new free_name_hash() used by read_cache.c::discard_index()
- remove_name_hash() takes an additional index_state parameter
- index_name_exists() for a directory (trailing '/') may return a cache
entry that has been removed (CE_UNHASHED). This is not a problem as the
return value is only used to check if the directory exists (dir.c) or to
normalize casing of directory names (read-cache.c).
Getting rid of cache_entry.dir_next reduces memory consumption, especially
with core.ignorecase=false (which doesn't use that member at all).
With core.ignorecase=true, building the directory index is slightly faster
as we add / check the parent directory first (instead of going through all
directory levels for each file in the index). E.g. with WebKit (~200k
files, ~7k dirs), time spent in lazy_init_name_hash is reduced from 176ms
to 130ms.
Signed-off-by: Karsten Blees <blees@dcon.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we can read packet data from memory as easily as a
descriptor, get_remote_heads can take either one as a
source. This will allow further refactoring in remote-curl.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
prepare_packed_git_one() is modified to allow count-objects to hook a
report function to so we don't need to duplicate the pack searching
logic in count-objects.c. When report_pack_garbage is NULL, the
overhead is insignificant.
The garbage is reported with warning() instead of error() in packed
garbage case because it's not an error to have garbage. Loose garbage
is still reported as errors and will be converted to warnings later.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Even though "git fetch" has full infrastructure to parse refspecs to
be fetched and match them against the list of refs to come up with
the final list of refs to be fetched, the list of refs that are
requested to be fetched were internally converted to a plain list of
strings at the transport layer and then passed to the underlying
fetch-pack driver.
Stop this conversion and instead pass around an array of refs.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Improve error and advice messages given locally when "git push"
refuses when it cannot compute fast-forwardness by separating these
cases from the normal "not a fast-forward; merge first and push
again" case.
* jc/push-reject-reasons:
push: finishing touches to explain REJECT_ALREADY_EXISTS better
push: introduce REJECT_FETCH_FIRST and REJECT_NEEDS_FORCE
push: further simplify the logic to assign rejection reason
push: further clean up fields of "struct ref"
Configuration parsing for tar.* configuration variables were
broken. Introduce a new config-keyname parser API to make the
callers much less error prone.
* jk/config-parsing-cleanup:
reflog: use parse_config_key in config callback
help: use parse_config_key for man config
submodule: simplify memory handling in config parsing
submodule: use parse_config_key when parsing config
userdiff: drop parse_driver function
convert some config callbacks to parse_config_key
archive-tar: use parse_config_key when parsing config
config: add helper function for parsing key names
Allow a configuration variable core.commentchar to customize the
character used to comment out the hint lines in the edited text from
the default '#'.
* jc/custom-comment-char:
Allow custom "comment char"
Some reimplementations of Git does not write all the stat info back
to the index due to their implementation limitations (e.g. jgit
running on Java). A configuration option can tell Git to ignore
changes to most of the stat fields and only pay attention to mtime
and size, which these implementations can reliably update. This
avoids excessive revalidation of contents.
* rr/minimal-stat:
Enable minimal stat checking
An element on GIT_CEILING_DIRECTORIES list that does not name the
real path to a directory (i.e. a symbolic link) could have caused
the GIT_DIR discovery logic to escape the ceiling.
* mh/ceiling:
string_list_longest_prefix(): remove function
setup_git_directory_gently_1(): resolve symlinks in ceiling paths
longest_ancestor_length(): require prefix list entries to be normalized
longest_ancestor_length(): take a string_list argument for prefixes
longest_ancestor_length(): use string_list_split()
Introduce new function real_path_if_valid()
real_path_internal(): add comment explaining use of cwd
Introduce new static function real_path_internal()
When we push to update an existing ref, if:
* the object at the tip of the remote is not a commit; or
* the object we are pushing is not a commit,
it won't be correct to suggest to fetch, integrate and push again,
as the old and new objects will not "merge". We should explain that
the push must be forced when there is a non-committish object is
involved in such a case.
If we do not have the current object at the tip of the remote, we do
not even know that object, when fetched, is something that can be
merged. In such a case, suggesting to pull first just like
non-fast-forward case may not be technically correct, but in
practice, most such failures are seen when you try to push your work
to a branch without knowing that somebody else already pushed to
update the same branch since you forked, so "pull first" would work
as a suggestion most of the time. And if the object at the tip is
not a commit, "pull first" will fail, without making any permanent
damage. As a side effect, it also makes the error message the user
will get during the next "push" attempt easier to understand, now
the user is aware that a non-commit object is involved.
In these cases, the current code already rejects such a push on the
client end, but we used the same error and advice messages as the
ones used when rejecting a non-fast-forward push, i.e. pull from
there and integrate before pushing again.
Introduce new rejection reasons and reword the messages
appropriately.
[jc: with help by Peff on message details]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "nonfastforward" and "update" fields are only used while
deciding what value to assign to the "status" locally in a single
function. Remove them from the "struct ref".
The "requires_force" field is not used to decide if the proposed
update requires a --force option to succeed, or to record such a
decision made elsewhere. It is used by status reporting code that
the particular update was "forced". Rename it to "forced_update",
and move the code to assign to it around to further clarify how it
is used and what it is used for.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Regression fix to stop "git push" complaining "target ref already
exists", when it is not the real reason the command rejected the
request (e.g. non-fast-forward).
* cr/push-force-tag-update:
push: fix "refs/tags/ hierarchy cannot be updated without --force"
The config callback functions get keys of the general form:
section.subsection.key
(where the subsection may be contain arbitrary data, or may
be missing). For matching keys without subsections, it is
simple enough to call "strcmp". Matching keys with
subsections is a little more complicated, and each callback
does it in an ad-hoc way, usually involving error-prone
pointer arithmetic.
Let's provide a helper that keeps the pointer arithmetic all
in one place.
Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Specifically the fields uid, gid, ctime, ino and dev are set to zero
by JGit. Other implementations, eg. Git in cygwin are allegedly also
somewhat incompatible with Git For Windows and on *nix platforms
the resolution of the timestamps may differ.
Any stat checking by git will then need to check content, which may
be very slow, particularly on Windows. Since mtime and size
is typically enough we should allow the user to tell git to avoid
checking these fields if they are set to zero in the index.
This change introduces a core.checkstat config option where the
the user can select to check all fields (default), or just size
and the whole second part of mtime (minimal).
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When pushing to update a branch with a commit that is not a
descendant of the commit at the tip, a wrong message "already
exists" was given, instead of the correct "non-fast-forward", if we
do not have the object sitting in the destination repository at the
tip of the ref we are updating.
The primary cause of the bug is that the check in a new helper
function is_forwardable() assumed both old and new objects are
available and can be checked, which is not always the case.
The way the caller uses the result of this function is also wrong.
If the helper says "we do not want to let this push go through", the
caller unconditionally translates it into "we blocked it because the
destination already exists", which is not true at all in this case.
Fix this by doing these three things:
* Remove unnecessary not_forwardable from "struct ref"; it is only
used inside set_ref_status_for_push();
* Make "refs/tags/" the only hierarchy that cannot be replaced
without --force;
* Remove the misguided attempt to force that everything that
updates an existing ref has to be a commit outside "refs/tags/"
hierarchy.
The policy last one tried to implement may later be resurrected and
extended to ensure fast-forwardness (defined as "not losing
objects", extending from the traditional "not losing commits from
the resulting history") when objects that are not commit are
involved (e.g. an annotated tag in hierarchies outside refs/tags),
but such a logic belongs to "is this a fast-forward?" check that is
done by ref_newer(); is_forwardable(), which is now removed, was not
the right place to do so.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some users do want to write a line that begin with a pound sign, #,
in their commit log message. Many tracking system recognise
a token of #<bugid> form, for example.
The support we offer these use cases is not very friendly to the end
users. They have a choice between
- Don't do it. Avoid such a line by rewrapping or indenting; and
- Use --cleanup=whitespace but remove all the hint lines we add.
Give them a way to set a custom comment char, e.g.
$ git -c core.commentchar="%" commit
so that they do not have to do either of the two workarounds.
[jc: although I started the topic, all the tests and documentation
updates, many of the call sites of the new strbuf_add_commented_*()
functions, and the change to git-submodule.sh scripted Porcelain are
from Ralf.]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Ralf Thielow <ralf.thielow@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit a469a10 wraps some error calls in macros to give the
compiler a chance to do more static analysis on their
constant -1 return value. We limit the use of these macros
to __GNUC__, since gcc is the primary beneficiary of the new
information, and because we use GNU features for handling
variadic macros.
However, clang also defines __GNUC__, but generates warnings
with -Wunused-value when these macros are used in a void
context, because the constant "-1" ends up being useless.
Gcc does not complain about this case (though it is unclear
if it is because it is smart enough to see what we are
doing, or too dumb to realize that the -1 is unused). We
can squelch the warning by just disabling these macros when
clang is in use.
Signed-off-by: Max Horn <max@quendi.de>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allow scripts to feed literal paths to commands that take
pathspecs, by disabling wildcard globbing.
* jk/pathspec-literal:
add global --literal-pathspecs option
Conflicts:
dir.c
Help compilers' flow analysis by making it more explicit that
error() always returns -1, to reduce false "variable used
uninitialized" warnings. Looks somewhat ugly but not too much.
* jk/error-const-return:
silence some -Wuninitialized false positives
make error()'s constant return value more visible
Allow us to read, and default to read, mailmap files from the tip
of the history in bare repositories. This will help running tools
like shortlog in server settings.
* jk/mailmap-from-blob:
mailmap: default mailmap.blob in bare repositories
mailmap: fix some documentation loose-ends for mailmap.blob
mailmap: clean up read_mailmap error handling
mailmap: support reading mailmap from blobs
mailmap: refactor mailmap parsing for non-file sources
Require "-f" for push to update a tag, even if it is a fast-forward.
* cr/push-force-tag-update:
push: allow already-exists advice to be disabled
push: rename config variable for more general use
push: cleanup push rules comment
push: clarify rejection of update to non-commit-ish
push: require force for annotated tags
push: require force for refs under refs/tags/
push: flag updates that require force
push: keep track of "update" state separately
push: add advice for rejected tag reference
push: return reject reasons as a bitset
Optimize matching paths with common forms of pathspecs that contain
wildcard characters.
* nd/pathspec-wildcard:
tree_entry_interesting: do basedir compare on wildcard patterns when possible
pathspec: apply "*.c" optimization from exclude
pathspec: do exact comparison on the leading non-wildcard part
pathspec: save the non-wildcard length part
An element on GIT_CEILING_DIRECTORIES list that does not name the
real path to a directory (i.e. a symbolic link) could have caused
the GIT_DIR discovery logic to escape the ceiling.
* mh/ceiling:
string_list_longest_prefix(): remove function
setup_git_directory_gently_1(): resolve symlinks in ceiling paths
longest_ancestor_length(): require prefix list entries to be normalized
longest_ancestor_length(): take a string_list argument for prefixes
longest_ancestor_length(): use string_list_split()
Introduce new function real_path_if_valid()
real_path_internal(): add comment explaining use of cwd
Introduce new static function real_path_internal()
Git takes pathspec arguments in many places to limit the
scope of an operation. These pathspecs are treated not as
literal paths, but as glob patterns that can be fed to
fnmatch. When a user is giving a specific pattern, this is a
nice feature.
However, when programatically providing pathspecs, it can be
a nuisance. For example, to find the latest revision which
modified "$foo", one can use "git rev-list -- $foo". But if
"$foo" contains glob characters (e.g., "f*"), it will
erroneously match more entries than desired. The caller
needs to quote the characters in $foo, and even then, the
results may not be exactly the same as with a literal
pathspec. For instance, the depth checks in
match_pathspec_depth do not kick in if we match via fnmatch.
This patch introduces a global command-line option (i.e.,
one for "git" itself, not for specific commands) to turn
this behavior off. It also has a matching environment
variable, which can make it easier if you are a script or
porcelain interface that is going to issue many such
commands.
This option cannot turn off globbing for particular
pathspecs. That could eventually be done with a ":(noglob)"
magic pathspec prefix. However, that level of granularity is
more cumbersome to use for many cases, and doing ":(noglob)"
right would mean converting the whole codebase to use
"struct pathspec", as the usual "const char **pathspec"
cannot represent extra per-item flags.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are a few error functions that simply wrap error() and
provide a standardized message text. Like error(), they
always return -1; knowing that can help the compiler silence
some false positive -Wuninitialized warnings.
One strategy would be to just declare these as inline in the
header file so that the compiler can see that they always
return -1. However, gcc does not always inline them (e.g.,
it will not inline opterror, even with -O3), which renders
our change pointless.
Instead, let's follow the same route we did with error() in
the last patch, and define a macro that makes the constant
return value obvious to the compiler.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a bare repository, there isn't a simple way to respect an
in-tree mailmap without extracting it to a temporary file.
This patch provides a config variable, similar to
mailmap.file, which reads the mailmap from a blob in the
repository.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
References are allowed to update from one commit-ish to another if the
former is an ancestor of the latter. This behavior is oriented to
branches which are expected to move with commits. Tag references are
expected to be static in a repository, though, thus an update to
something under refs/tags/ should be rejected unless the update is
forced.
Signed-off-by: Chris Rorvick <chris@rorvick.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a flag for indicating an update to a reference requires force.
Currently the `nonfastforward` flag is used for this when generating the
status message. A separate flag insulates dependent logic from the
details of set_ref_status_for_push().
Signed-off-by: Chris Rorvick <chris@rorvick.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the reference exists on the remote and it is not being removed, then
mark as an update. This is in preparation for handling tags (lightweight
and annotated) exceptionally.
Signed-off-by: Chris Rorvick <chris@rorvick.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Advising the user to fetch and merge only makes sense if the rejected
reference is a branch. If none of the rejections are for branches, just
tell the user the reference already exists.
Signed-off-by: Chris Rorvick <chris@rorvick.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a pattern contains only a single asterisk as wildcard,
e.g. "foo*bar", after literally comparing the leading part "foo" with
the string, we can compare the tail of the string and make sure it
matches "bar", instead of running fnmatch() on "*bar" against the
remainder of the string.
-O2 build on linux-2.6, without the patch:
$ time git rev-list --quiet HEAD -- '*.c'
real 0m40.770s
user 0m40.290s
sys 0m0.256s
With the patch
$ time ~/w/git/git rev-list --quiet HEAD -- '*.c'
real 0m34.288s
user 0m33.997s
sys 0m0.205s
The above command is not supposed to be widely popular. It's chosen
because it exercises pathspec matching a lot. The point is it cuts
down matching time for popular patterns like *.c, which could be used
as pathspec in other places.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We mark pathspec with wildcards with the field use_wildcard. We
could do better by saving the length of the non-wildcard part, which
can be used for optimizations such as f9f6e2c (exclude: do strcmp as
much as possible before fnmatch - 2012-06-07).
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We keep track of whether the user ident was given to us
explicitly, or if we guessed at it from system parameters
like username and hostname. However, we kept only a single
variable. This covers the common cases (because the author
and committer will usually come from the same explicit
source), but can miss two cases:
1. GIT_COMMITTER_* is set explicitly, but we fallback for
GIT_AUTHOR. We claim the ident is explicit, even though
the author is not.
2. GIT_AUTHOR_* is set and we ask for author ident, but
not committer ident. We will claim the ident is
implicit, even though it is explicit.
This patch uses two variables instead of one, updates both
when we set the "fallback" values, and updates them
individually when we read from the environment.
Rather than keep user_ident_sufficiently_given as a
compatibility wrapper, we update the only two callers to
check the committer_ident, which matches their intent and
what was happening already.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In v1.5.6-rc0~56^2 (2008-05-04) "user_ident_explicitly_given"
was introduced as a global for communication between config,
ident, and builtin-commit. In v1.7.0-rc0~72^2 (2010-01-07)
readers switched to using the common wrapper
user_ident_sufficiently_given(). After v1.7.11-rc1~15^2~18
(2012-05-21), the var is only written in ident.c.
Now we can make it static, which will enable further
refactoring without worrying about upsetting other code.
Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is used by diff-no-index.c, part of libgit.a while it stays in
builtin/diff.c. Move it to diff.c so that we won't get undefined
reference if a program that uses libgit.a happens to pull it in.
While at it, move check_pager from git.c to pager.c. It makes more
sense there and pager.c is also part of libgit.a
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
These functions are called in sequencer.c, which is part of
libgit.a. This makes libgit.a potentially require builtin/merge.c for
external git commands.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Change longest_ancestor_length() to take the prefixes argument as a
string_list rather than as a colon-separated string. This will make
it easier for the caller to alter the entries before calling
longest_ancestor_length().
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
The function is like real_path(), except that it returns NULL on error
instead of dying.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
The exit status code from "git config" was way overspecified while
being incorrect. Update the implementation to give the documented
status for a case that was documented, and introduce a new code for
"all other errors".
* jc/maint-config-exit-status:
config: "git config baa" should exit with status 1
Some capabilities were asked by fetch-pack even when upload-pack did
not advertise that they are available. Fix fetch-pack not to do so.
* jc/capabilities:
fetch-pack: mention server version with verbose output
parse_feature_request: make it easier to see feature values
fetch-pack: do not ask for unadvertised capabilities
do not send client agent unless server does first
send-pack: fix capability-sending logic
include agent identifier in capability string
We already take care to parse key/value capabilities like
"foo=bar", but the code does not provide a good way of
actually finding out what is on the right-hand side of the
"=".
A server using "parse_feature_request" could accomplish this
with some extra parsing. You must skip past the "key"
portion manually, check for "=" versus NUL or space, and
then find the length by searching for the next space (or
NUL). But clients can't even do that, since the
"server_supports" interface does not even return the
pointer.
Instead, let's have our parser share more information by
providing a pointer to the value and its length. The
"parse_feature_value" function returns a pointer to the
feature's value portion, along with the length of the value.
If the feature is missing, NULL is returned. If it does not
have an "=", then a zero-length value is returned.
Similarly, "server_feature_value" behaves in the same way,
but always checks the static server_feature_list variable.
We can then implement "server_supports" in terms of
"server_feature_value". We cannot implement the original
"parse_feature_request" in terms of our new function,
because it returned a pointer to the beginning of the
feature. However, no callers actually cared about the value
of the returned pointer, so we can simplify it to a boolean
just as we do for "server_supports".
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We instead failed with an undocumented exit status 255.
Also define a "catch-all" status and document it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Split lower bits of ce_flags field and creates a new ce_namelen
field in the in-core index structure.
* tg/ce-namelen-field:
Strip namelen out of ce_flags into a ce_namelen field
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"
...
Teaches git to normalize pathnames read from readdir(3) and all
arguments from the command line into precomposed UTF-8 (assuming
that they come as decomposed UTF-8) to work around issues on Mac OS.
I think there still are other places that need conversion
(e.g. paths that are read from stdin for some commands), but this
should be a good first step in the right direction.
* tb/sanitize-decomposed-utf-8-pathname:
git on Mac OS and precomposed unicode
Strip the name length from the ce_flags field and move it
into its own ce_namelen field in struct cache_entry. This
will both give us a tiny bit of a performance enhancement
when working with long pathnames and is a refactoring for
more readability of the code.
It enhances readability, by making it more clear what
is a flag, and where the length is stored and make it clear
which functions use stages in comparisions and which only
use the length.
It also makes CE_NAMEMASK private, so that users don't
mistakenly write the name length in the flags.
Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new option allows you to feed an ambiguous prefix and enumerate
all the objects that share it as a prefix of their object names.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This teaches the revision parser that in "$name:$path" (used for a
blob object name), "$name" must be a tree-ish.
There are many more places where we know what types of objects are
called for. This patch adds support for "commit", "treeish", "tree",
and "blob", which could be used in the following contexts:
- "git apply --build-fake-ancestor" reads the "index" lines from
the patch; they must name blob objects (not even "blob-ish");
- "git commit-tree" reads a tree object name (not "tree-ish"), and
zero or more commit object names (not "committish");
- "git reset $rev" wants a committish; "git reset $rev -- $path"
wants a treeish.
They will come in later patches in the series.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many callers know that the user meant to name a committish by
syntactical positions where the object name appears. Calling this
function allows the machinery to disambiguate shorter-than-unique
abbreviated object names between committish and others.
Note that this does NOT error out when the named object is not a
committish. It is merely to give a hint to the disambiguation
machinery.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function takes user input string and returns the object name
(binary SHA-1) with mode bits and path when the object was looked
up in a tree.
Additionally give hints to help disambiguation of abbreviated object
names when the caller knows what it is looking for.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We know that the token "$name" that appear in "$name^{commit}",
"$name^4", "$name~4" etc. can only name a committish (either a
commit or a tag that peels to a commit). Teach get_short_sha1() to
take advantage of that knowledge when disambiguating an abbreviated
SHA-1 given as an object name.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach git to read various information from $XDG_CONFIG_HOME/git/ to allow
the user to avoid cluttering $HOME.
* mm/config-xdg:
config: write to $XDG_CONFIG_HOME/git/config file when appropriate
Let core.attributesfile default to $XDG_CONFIG_HOME/git/attributes
Let core.excludesfile default to $XDG_CONFIG_HOME/git/ignore
config: read (but not write) from $XDG_CONFIG_HOME/git/config file
Mac OS X mangles file names containing unicode on file systems HFS+,
VFAT or SAMBA. When a file using unicode code points outside ASCII
is created on a HFS+ drive, the file name is converted into
decomposed unicode and written to disk. No conversion is done if
the file name is already decomposed unicode.
Calling open("\xc3\x84", ...) with a precomposed "Ä" yields the same
result as open("\x41\xcc\x88",...) with a decomposed "Ä".
As a consequence, readdir() returns the file names in decomposed
unicode, even if the user expects precomposed unicode. Unlike on
HFS+, Mac OS X stores files on a VFAT drive (e.g. an USB drive) in
precomposed unicode, but readdir() still returns file names in
decomposed unicode. When a git repository is stored on a network
share using SAMBA, file names are send over the wire and written to
disk on the remote system in precomposed unicode, but Mac OS X
readdir() returns decomposed unicode to be compatible with its
behaviour on HFS+ and VFAT.
The unicode decomposition causes many problems:
- The names "git add" and other commands get from the end user may
often be precomposed form (the decomposed form is not easily input
from the keyboard), but when the commands read from the filesystem
to see what it is going to update the index with already is on the
filesystem, readdir() will give decomposed form, which is different.
- Similarly "git log", "git mv" and all other commands that need to
compare pathnames found on the command line (often but not always
precomposed form; a command line input resulting from globbing may
be in decomposed) with pathnames found in the tree objects (should
be precomposed form to be compatible with other systems and for
consistency in general).
- The same for names stored in the index, which should be
precomposed, that may need to be compared with the names read from
readdir().
NFS mounted from Linux is fully transparent and does not suffer from
the above.
As Mac OS X treats precomposed and decomposed file names as equal,
we can
- wrap readdir() on Mac OS X to return the precomposed form, and
- normalize decomposed form given from the command line also to the
precomposed form,
to ensure that all pathnames used in Git are always in the
precomposed form. This behaviour can be requested by setting
"core.precomposedunicode" configuration variable to true.
The code in compat/precomposed_utf8.c implements basically 4 new
functions: precomposed_utf8_opendir(), precomposed_utf8_readdir(),
precomposed_utf8_closedir() and precompose_argv(). The first three
are to wrap opendir(3), readdir(3), and closedir(3) functions.
The argv[] conversion allows to use the TAB filename completion done
by the shell on command line. It tolerates other tools which use
readdir() to feed decomposed file names into git.
When creating a new git repository with "git init" or "git clone",
"core.precomposedunicode" will be set "false".
The user needs to activate this feature manually. She typically
sets core.precomposedunicode to "true" on HFS and VFAT, or file
systems mounted via SAMBA.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Torsten Bögershausen <tboegi@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git diff --no-index" did not correctly handle relative paths and
did not give correct exit codes when run under "--quiet" option.
* th/diff-no-index-fixes:
diff-no-index: exit(1) if 'diff --quiet <repo file> <external file>' finds changes
diff: handle relative paths in no-index
