This is a long overdue fix to the API for for_each_ref() family
of functions. It allows the callers to specify a callback data
pointer, so that the caller does not have to use static
variables to communicate with the callback funciton.
The updated for_each_ref() family takes a function of type
int (*fn)(const char *, const unsigned char *, void *)
and a void pointer as parameters, and calls the function with
the name of the ref and its SHA-1 with the caller-supplied void
pointer as parameters.
The commit updates two callers, builtin-name-rev.c and
builtin-pack-refs.c as an example.
An earlier conversion accidentally hardcoded "HEAD" to be passed to
resolve_ref(), thereby causing git-symbolic-ref command to always
report where the HEAD points at, ignoring the command line parameter.
Current git#next is totally broken wrt. cloning over HTTP, generating refs
at random directories. Of course it's caused by the static get_pathname()
buffer. lock_ref_sha1() stores return value of mkpath()'s get_pathname()
call, then calls lock_ref_sha1_basic() which calls git_path(ref) which
calls get_pathname() at that point returning pointer to the same buffer.
So now you are sprintf()ing a format string into itself, wow! The resulting
pathnames are really cute. (If you've been paying attention, yes, the
mere fact that a format string _could_ write over itself is very wrong
and probably exploitable here. See the other mail I've just sent.)
I've never liked how we use return values of those functions so liberally,
the "allow some random number of get_pathname() return values to work
concurrently" is absolutely horrible pit and we've already fallen in this
before IIRC. I consider it an awful coding practice, you add a call
somewhere and at some other point some distant caller of that breaks since
it reuses the same return values. Not to mention this takes quite some time
to debug.
My gut feeling tells me that there might be more of this. I don't have
time to review the rest of the users of the refs.c functions though.
Signed-off-by: Petr Baudis <pasky@suse.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
This actually "turns on" the packed ref file format, now that the
infrastructure to do so sanely exists (ie notably the change to make the
reference reading logic take refnames rather than pathnames to the loose
objects that no longer necessarily even exist).
In particular, when the ref lookup hits a refname that has no loose file
associated with it, it falls back on the packed-ref information. Also, the
ref-locking code, while still using a loose file for the locking itself
(and _creating_ a loose file for the new ref) no longer requires that the
old ref be in such an unpacked state.
Finally, this does a minimal hack to git-checkout.sh to rather than check
the ref-file directly, do a "git-rev-parse" on the "heads/$refname".
That's not really wonderful - we should rather really have a special
routine to verify the names as proper branch head names, but it is a
workable solution for now.
and the end result is a largely working repository (ie I've done two
commits - which creates _one_ unpacked ref file - done things like run
"gitk" and "git log" etc, and it all looks ok).
There are probably things missing, but I'm hoping that the missing things
are now of the "small and obvious" kind, and that somebody else might want
to start looking at this too. Hint hint ;)
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
The old code used to totally mix up the notion of a ref-name and the path
that that ref was associated with. That was not only horribly ugly (a
number of users got the path, and then wanted to try to turn it back into
a ref-name again), but it fundamnetally doesn't work at all once we do any
setup where a ref doesn't have a 1:1 relationship with a particular
pathname.
This fixes things up so that we use the ref-name throughout, and only
turn it into a pathname once we actually look it up in the filesystem.
That makes a lot of things much clearer and more straightforward.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
You can remove a ref that is packed two different ways: either simply
repack all the refs without that one, or create a loose ref that has the
magic all-zero SHA1.
This also adds back the test that a ref actually has the object it
points to.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Start handling references internally as a sorted in-memory list
This also adds some very rudimentary support for the notion of packed
refs. HOWEVER! At this point it isn't used to actually look up a ref
yet, only for listing them (ie "for_each_ref()" and friends see the
packed refs, but none of the other single-ref lookup routines).
Note how we keep two separate lists: one for the loose refs, and one for
the packed refs we read. That's so that we can easily keep the two apart,
and read only one set or the other (and still always make sure that the
loose refs take precedence).
[ From this, it's not actually obvious why we'd keep the two separate
lists, but it's important to have the packed refs on their own list
later on, when I add support for looking up a single loose one.
For that case, we will want to read _just_ the packed refs in case the
single-ref lookup fails, yet we may end up needing the other list at
some point in the future, so keeping them separated is important ]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
* jc/pack:
pack-objects: document --revs, --unpacked and --all.
pack-objects --unpacked=<existing pack> option.
pack-objects: further work on internal rev-list logic.
pack-objects: run rev-list equivalent internally.
Separate object listing routines out of rev-list
* jk/diff:
wt-status: remove extraneous newline from 'deleted:' output
git-status: document colorization config options
Teach runstatus about --untracked
git-commit.sh: convert run_status to a C builtin
Move color option parsing out of diff.c and into color.[ch]
diff: support custom callbacks for output
upload-archive: monitor child communication even more carefully.
The current code works like this: if others flags than POLLIN is
raised we assume that (a) something bad happened and the child died or
(b) the child has closed the pipe because it had no more data to send.
For the latter case, we assume wrongly that one call to
process_input() will empty the pipe. Indeed it reads only 16Ko of data
by call and the the pipe capacity can be larger than that (on current
Linux kernel, it is 65536 bytes). Therefore the child can write 32ko
of data, for example, and close the pipe. After that poll will return
POLLIN _and_ POLLHUP and the parent will read only 16ko of data.
This patch forces the parent to empty the pipe as soon as POLLIN is
raised and even if POLLHUP or something else is raised too.
Moreover, some implementations of poll might return POLLRDNORM flag
even if it is non standard.
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
* jc/archive:
git-tar-tree: devolve git-tar-tree into a wrapper for git-archive
git-archive: inline default_parse_extra()
builtin-archive.c: rename remote_request() to extract_remote_arg()
upload-archive: monitor child communication more carefully.
Add sideband status report to git-archive protocol
Prepare larger packet buffer for upload-pack protocol.
Teach --exec to git-archive --remote
Add --verbose to git-archive
archive: force line buffered output to stderr
Use xstrdup instead of strdup in builtin-{tar,zip}-tree.c
Move sideband server side support into reusable form.
Move sideband client side support into reusable form.
archive: allow remote to have more formats than we understand.
git-archive: make compression level of ZIP archives configurable
Add git-upload-archive
git-archive: wire up ZIP format.
git-archive: wire up TAR format.
Add git-archive
$export_ok: If this variable evaluates to true it is checked
if a file with this name exists in the repository. If it
does not exist the repository cannot be viewed from gitweb.
(Similar to git-daemon-export-ok for git-daemon).
$strict_export: If this variable evaluates to true only
repositories listed on the project-list-page of gitweb can
be accessed.
git-tar-tree: devolve git-tar-tree into a wrapper for git-archive
This patch removes the custom tree walker tree_traverse(), and makes
generate_tar() use write_tar_archive() and the infrastructure provided
by git-archive instead.
As a kind of side effect, make write_tar_archive() able to handle NULL
as base directory, as this is what the new and simple generate_tar()
uses to indicate the absence of a base directory. This was simpler
and cleaner than playing tricks with empty strings.
The behaviour of git-tar-tree should be unchanged (quick tests didn't
indicate otherwise) except for the text of some error messages.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
Now three types of path based URLs are supported:
gitweb.cgi/project.git
gitweb.cgi/project.git/branch
gitweb.cgi/project.git/branch/filename
The first one (show project summary) was already supported for a long time
now. The other two are new: they show the shortlog of a branch or
the plain file contents of some file contained in the repository.
This is especially useful to support project web pages for small
projects: just create an html branch and then use an URL like
gitweb.cgi/project.git/html/index.html.
Signed-off-by: Martin Waitz <tali@admingilde.org> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
apply --unidiff-zero: loosen sanity checks for --unidiff=0 patches
In "git-apply", we have a few sanity checks and heuristics that
expects that the patch fed to us is a unified diff with at least
one line of context.
* When there is no leading context line in a hunk, the hunk
must apply at the beginning of the preimage. Similarly, no
trailing context means that the hunk is anchored at the end.
* We learn a patch deletes the file from a hunk that has no
resulting line (i.e. all lines are prefixed with '-') if it
has not otherwise been known if the patch deletes the file.
Similarly, no old line means the file is being created.
And we declare an error condition when the file created by a
creation patch already exists, and/or when a deletion patch
still leaves content in the file.
These sanity checks are good safety measures, but breaks down
when people feed a diff generated with --unified=0. This was
recently noticed first by Matthew Wilcox and Gerrit Pape.
This adds a new flag, --unified-zero, to allow bypassing these
checks. If you are in control of the patch generation process,
you should not use --unified=0 patch and fix it up with this
flag; rather you should try work with a patch with context. But
if all you have to work with is a patch without context, this
flag may come handy as the last resort.
I had a hard time figuring out why this test was failing with
the packed-refs update without running it under "sh -x". This
makes output from "sh t1400-update-ref.sh -v" more descriptive.
Updating other tests would be a good janitorial task.
http-fetch.c: consolidate code to detect missing fetch target
At a handful places we check two error codes from curl library
to see if the file we asked was missing from the remote (e.g.
we asked for a loose object when it is in a pack) to decide what
to do next. This consolidates the check into a single function.
NOTE: the original did not check for HTTP_RETURNED_ERROR when
error code is 404, but this version does to make sure 404 is
from HTTP and not some other protcol.
An earlier commit cbd64af added a check that prevents "git-am"
to run without its standard input connected to a terminal while
resuming operation. This was to catch a user error to try
feeding a new patch from its standard input while recovery.
The assumption of the check was that it is an indication that a
new patch is being fed if the standard input is not connected to
a terminal. It is however not quite correct (the standard input
can be /dev/null if the user knows the operation does not need
any input, for example). This broke t3403 when the test was run
with its standard input connected to /dev/null.
When git-am is given an explicit command such as --skip, there
is no reason to insist that the standard input is a terminal; we
are not going to read a new patch anyway.
Credit goes to Gerrit Pape for noticing and reporting the
problem with t3403-rebase-skip test.
This allows you to maintain a few filesystem pathnames concurrently, by
simply replacing the single static "pathname" buffer with a LRU of four
buffers.
We did exactly the same thing with sha1_to_hex(), for pretty much exactly
the same reason. Sometimes you want to use two pathnames, and while it's
easy enough to xstrdup() them, why not just do the LU buffer thing.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
gitweb: Allow for href() to be used for links without project param
Make it possible to use href() subroutine to generate link with
query string which does not include project ('p') parameter.
href() used to add project=$project to its parameters, if it
was not set (to be more exact if $params{'project'} was false).
Now you can pass "project => undef" if you don't want for href()
to add project parameter to query string in the generated link.
Links to "project_list", "project_index" and "opml" (all related
to list of all projects/all git repositories) doesn't need project
parameter. Moreover "project_list" is default view (action) if
project ('p') parameter is not set, just like "summary" is default
view (action) if project is set; project list served as a kind
of "home" page for gitweb instalation, and links to "project_list"
view were done without specyfying it as an action.
Convert remaining links (except $home_link and anchor links)
to use href(); this required adding 'order => "o"' to @mapping
in href(). This finishes consolidation of URL generation.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
gitweb: Add git_project_index for generating index.aux
Add git_project_index, which generates index.aux file that can be used
as a source of projects list, instead of generating projects list from
a directory. Using file as a source of projects list allows for some
projects to be not present in gitweb main (project_list) page, and/or
correct project owner info. And is probably faster.
Additionally it can be used to get the list of all available repositories
for scripts (in easily parseable form).
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
gitweb: Do not parse refs by hand, use git-peek-remote instead
This is in response to Linus's work on packed refs. Additionally it
makes gitweb work with symrefs, too.
Do not parse refs by hand, using File::Find and reading individual
heads to get hash of reference, but use git-peek-remote output
instead. Assume that the hash for deref (with ^{}) always follows hash
for ref, and that we have derefs only for tag objects; this removes
call to git_get_type (and git-cat-file -t invocation) for tags, which
speeds "summary" and "tags" views generation, but might slow generation
of "heads" view a bit. For now, we do not save and use the deref hash.
Remove git_get_hash_by_ref while at it, as git_get_refs_list was the
only place it was used.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
gitweb: Use File::Find::find in git_get_projects_list
Earlier code to get list of projects when $projects_list is a
directory (e.g. when it is equal to $projectroot) had a hardcoded flat
(one level) list of directories. Allow for projects to be in
subdirectories also for $projects_list being a directory by using
File::Find.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
It turns out that I actually wanted to avoid the filenames (because I
didn't care - I just wanted to see the context in which something was
used) when doing a grep. But since "git grep" didn't take the "-h"
parameter, I ended up having to do "grep -5 -h *.c" instead.
So here's a trivial patch that adds "-h" (and thus has to enable -H too)
to "git grep" parsing.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Fetch over http from a repository that uses alternates to borrow
from neighbouring repositories were quite broken, apparently for
some time now.
We parse input and count bytes to allocate the new buffer, and
when we copy into that buffer we know exactly how many bytes we
want to copy from where. Using strlcpy for it was simply
stupid, and the code forgot to take it into account that strlcpy
terminated the string with NUL.
Fetch over http from a repository that uses alternates to borrow
from neighbouring repositories were quite broken, apparently for
some time now.
We parse input and count bytes to allocate the new buffer, and
when we copy into that buffer we know exactly how many bytes we
want to copy from where. Using strlcpy for it was simply
stupid, and the code forgot to take it into account that strlcpy
terminated the string with NUL.
Actually, teach runstatus what to do if it is not passed; it should not list
the contents of completely untracked directories, but only the name of that
directory (plus a trailing '/').
[jc: with comments by Jeff King to match hide-empty-directories
behaviour of the original.]
Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
upload-archive: monitor child communication more carefully.
Franck noticed that the code around polling and relaying messages
from the child process was quite bogus. Here is an attempt to
clean it up a bit, based on his patch:
- When POLLHUP is set, it goes ahead and reads the file
descriptor. Worse yet, it does not check the return value of
read() for errors when it does.
- When we processed one POLLIN, we should just go back and see
if any more data is available. We can check if the child is
still there when poll gave control back at us but without any
actual input.
git_connect() can return 0 if we use git protocol for example.
Users of this function don't know and don't care if a process
had been created or not, and to avoid them to check it before
calling finish_connect() this patch allows finish_connect() to
take a null pid. And in that case return 0.
[jc: updated function signature of git_connect() with a comment on
its return value. ]
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
Fix a memory leak in "connect.c" and die if command too long.
Use "add_to_string" instead of "sq_quote" and "snprintf", so
that there is no memory allocation and no memory leak.
Also check if the command is too long to fit into the buffer
and die if this is the case, instead of truncating it to the
buffer size.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
git_history output is now divided into pages, like git_shortlog,
git_tags and git_heads output. As whole git-rev-list output is now
read into array before writing anything, it allows for better
signaling of errors.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
As pickaxe search (selected using undocumented 'pickaxe:' operator in
search query) is resource consuming, allow to turn it on/off using
feature meachanism. Turned on by default, for historical reasons.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
Add sideband status report to git-archive protocol
Using the refactored sideband code from existing upload-pack protocol,
this lets the error condition and status output sent from the remote
process to be shown locally.
* jc/sideband:
Prepare larger packet buffer for upload-pack protocol.
Move sideband server side support into reusable form.
Move sideband client side support into reusable form.
get_sha1_hex() micro-optimization
Prepare larger packet buffer for upload-pack protocol.
The original side-band support added to the upload-pack protocol used the
default 1000-byte packet length. The pkt-line format allows up to 64k, so
prepare the receiver for the maximum size, and have the uploader and
downloader negotiate if larger packet length is allowed.
Some people needed --exec to specify the location of the upload-pack
executable, because their default SSH log-in does not include the
directory they have their own private copy of git on the $PATH.
These people need to be able to say --exec to git-archive --remote
for the same reason.
Use xstrdup instead of strdup in builtin-{tar,zip}-tree.c
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
(cherry picked from 5d2aea4cb383a43e40d47ab69d8ad7a495df6ea2 commit)
archive: allow remote to have more formats than we understand.
This fixes git-archive --remote not to parse archiver arguments;
otherwise if the remote end implements formats other than the
one known locally we will not be able to access that format.
This command implements the git archive protocol on the server
side. This command is not intended to be used by the end user.
Underlying git-archive command line options are sent over the
protocol from "git-archive --remote=...", just like upload-tar
currently does with "git-tar-tree=...".
As for "git-archive" command implementation, this new command
does not execute any existing "git-{tar,zip}-tree" but rely
on the archive API defined by "git-archive" patch. Hence we
get 2 good points:
- "git-archive" and "git-upload-archive" share all option
parsing code.
- All kind of git-upload-{tar,zip} can be deprecated.
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-archive is a command to make TAR and ZIP archives of a git tree.
It helps prevent a proliferation of git-{format}-tree commands.
Instead of directly calling git-{tar,zip}-tree command, it defines
a very simple API, that archiver should implement and register in
"git-archive.c". This API is made up by 2 functions whose prototype
is defined in "archive.h" file.
- The first one is used to parse 'extra' parameters which have
signification only for the specific archiver. That would allow
different archive backends to have different kind of options.
- The second one is used to ask to an archive backend to build
the archive given some already resolved parameters.
The main reason for making this API is to avoid using
git-{tar,zip}-tree commands, hence making them useless. Maybe it's
time for them to die ?
It also implements remote operations by defining a very simple
protocol: it first sends the name of the specific uploader followed
the repository name (git-upload-tar git://example.org/repo.git).
Then it sends options. It's done by sending a sequence of one
argument per packet, with prefix "argument ", followed by a flush.
The remote protocol is implemented in "git-archive.c" for client
side and is triggered by "--remote=<repo>" option. For example,
to fetch a TAR archive in a remote repo, you can issue:
$ git archive --format=tar --remote=git://xxx/yyy/zzz.git HEAD
We choose to not make a new command "git-fetch-archive" for example,
avoind one more GIT command which should be nice for users (less
commands to remember, keeps existing --remote option).
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com> Acked-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <junkio@cox.net>
This creates a new git-runstatus which should do roughly the same thing
as the run_status function from git-commit.sh. Except for color support,
the main focus has been to keep the output identical, so that it can be
verified as correct and then used as a C platform for other improvements to
the status printing code.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <junkio@cox.net>
autoconf: Add support for setting NO_ICONV and ICONVDIR
Add support for ./configure options --without-iconv (if neither libc
nor libiconv properly support iconv), and for --with-iconv=PATH (to
set prefix to libiconv library and headers, used only when
NEED_LIBICONV is set). While at it, make ./configure set or unset
NO_ICONV always (it is not autodetected in Makefile).
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
which picks up all loose objects that are still live and creates
a new pack.
This implements --unpacked=<existing pack> option to tell the
revision walking machinery to pretend as if objects in such a
pack are unpacked for the purpose of object listing. With this,
we could say:
instead, to mean "all live loose objects but pretend as if
objects that are in this pack are also unpacked". The newly
created pack would be perfect for updating $active_pack by
replacing it.
Since pack-objects now knows how to do the rev-list's work
itself internally, you can also write the above example by:
Historically we did not allow binary patch applied without an
explicit permission from the user, and this flag was the way to
do so. This makes the flag a no-op by always allowing binary
patch application.
When we are generating packs to update remote repositories we
want to supply as much information as possible about the revisions
that already exist to rev-list in order optimise the pack as much
as possible. We need to pass two revisions for each branch we are
updating in the remote repository and one for each additional branch.
Where the remote repository has numerous branches we can run out
of command line space to pass them.
Utilise the git-rev-list --stdin mode to allow unlimited numbers
of revision constraints. This allows us to move back to the much
simpler unordered revision selection code.
[jc: added some comments in the code to describe the pipe flow
a bit.]
Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Junio C Hamano <junkio@cox.net>