git.git
13 years agoMakefile: list one vcs-svn/xdiff object or header per line
Jonathan Nieder [Sun, 27 Mar 2011 18:13:22 +0000 (13:13 -0500)]
Makefile: list one vcs-svn/xdiff object or header per line

As the svn import infrastructure evolves, it's getting to be a pain to
tell by eye what files were added or removed from a dependency line
like

VCSSVN_OBJS = vcs-svn/string_pool.o vcs-svn/line_buffer.o \
vcs-svn/repo_tree.o vcs-svn/fast_export.o vcs-svn/svndump.o

So use a style with one entry per line instead, like the existing
BUILTIN_OBJS:

# protect against environment
VCSSVN_OBJS =
...
VCSSVN_OBJS += vcs-svn/string_pool.o
VCSSVN_OBJS += vcs-svn/line_buffer.o
...

which is readable on its own and produces nice, clear diffs.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: add missing cast to printf argument
Jonathan Nieder [Sun, 27 Mar 2011 17:19:14 +0000 (12:19 -0500)]
vcs-svn: add missing cast to printf argument

gcc -m32 correctly warns:

 vcs-svn/fast_export.c: In function 'fast_export_commit':
 vcs-svn/fast_export.c:54:2: warning: format '%llu' expects
   argument of type 'long long unsigned int', but argument 2
   has type 'unsigned int' [-Wformat]

Fix it.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: handle log message with embedded NUL
Jonathan Nieder [Sat, 26 Mar 2011 05:49:37 +0000 (00:49 -0500)]
vcs-svn: handle log message with embedded NUL

Pass the log message by strbuf instead of as a C-style string and use
fwrite instead of printf to write it to fast-import so embedded '\0'
bytes can be preserved.

Currently "git log" doesn't show the embedded NULs but "git cat-file
commit" can.

While at it, stop including system headers from repo_tree.h.  git
source files need to include git-compat-util.h (or cache.h or
builtin.h) sooner to ensure the appropriate feature test macros are
defined.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: avoid unnecessary copying of log message and author
Jonathan Nieder [Fri, 25 Mar 2011 04:10:00 +0000 (23:10 -0500)]
vcs-svn: avoid unnecessary copying of log message and author

Use strbuf_swap when storing the svn:log and svn:author properties, so
pointers to rather than the contents of buffers get copied.  The main
effect should be to make the code a little easier to read.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: remove buffer_read_string
Jonathan Nieder [Fri, 25 Mar 2011 04:09:19 +0000 (23:09 -0500)]
vcs-svn: remove buffer_read_string

All previous users of buffer_read_string have already been converted
to use the more intuitive buffer_read_binary, so remove the old API to
avoid some confusion.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: make reading of properties binary-safe
Jonathan Nieder [Sat, 26 Mar 2011 05:15:10 +0000 (00:15 -0500)]
vcs-svn: make reading of properties binary-safe

svn-fe errors out on revision 59151 of the ASF repository:

 fatal: invalid dump: unexpected end of file

The proximate cause is a property with an embedded NUL character.
Previously such anomalies were ignored but commit c9d1c8ba
(2010-12-28) introduced a check strlen(val) == len to avoid reading
uninitialized data when a property list ends early and unfortunately
this test does not distinguish between "foo" followed by EOF and the
string "foo\0bar\0baz".

Fix it by using buffer_read_binary to read to a strbuf and checking
the actual length read.  Most consumers of properties still use
C-style strings, so in practice an author or log message with embedded
NULs will be truncated, but a least this way svn-fe won't error out
(fixing the regression).

Reported-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agoMerge branch 'db/length-as-hash' into svn-fe
Jonathan Nieder [Tue, 22 Mar 2011 23:44:49 +0000 (18:44 -0500)]
Merge branch 'db/length-as-hash' into svn-fe

* db/length-as-hash:
  vcs-svn: use strchr to find RFC822 delimiter
  vcs-svn: implement perfect hash for top-level keys
  vcs-svn: implement perfect hash for node-prop keys

Conflicts:
vcs-svn/svndump.c

13 years agovcs-svn: use strchr to find RFC822 delimiter
David Barr [Tue, 14 Dec 2010 00:06:43 +0000 (11:06 +1100)]
vcs-svn: use strchr to find RFC822 delimiter

This is a small optimisation (4% reduction in user time) but is the
largest artifact within the parsing portion of svndump.c

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: implement perfect hash for top-level keys
David Barr [Mon, 13 Dec 2010 08:56:01 +0000 (19:56 +1100)]
vcs-svn: implement perfect hash for top-level keys

Instead of interning property names and comparing their string_pool
keys, look them up in a table by string length, which should be about
as fast.

Another small step towards removing dependence on string_pool
altogether.

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: implement perfect hash for node-prop keys
David Barr [Mon, 13 Dec 2010 08:13:24 +0000 (19:13 +1100)]
vcs-svn: implement perfect hash for node-prop keys

Instead of interning property names and comparing their string_pool
keys, look them up in a table by string length, which should be about
as fast.

This is a small step towards removing dependence on string_pool.

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: use strbuf for author, UUID, and URL
David Barr [Tue, 22 Mar 2011 22:52:17 +0000 (17:52 -0500)]
vcs-svn: use strbuf for author, UUID, and URL

Use strbufs and strings instead of interned strings for values of rev,
dump, and node fields that happen to be strings.  After this change,
the only remaining string_pool use is for paths in the repo_tree API
and internals.

Functional change: treat an empty author, UUID, or URL as none at all.
So for example, in repos where the first revision has an empty
svn:author property, the first rev will be treated as by "nobody"
rather than by a person with empty name and email address created by
prepending an @ sign to the repository UUID.

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: use strbuf for revision log
David Barr [Mon, 21 Mar 2011 23:49:50 +0000 (10:49 +1100)]
vcs-svn: use strbuf for revision log

obj_pool is overkill for this application: all that is needed is a
buffer that can resize from rev to rev to accomodate differently-sized
strings.  In the spirit of commit deadcef4 (2010-11-06), use a strbuf
instead.

This is a small step towards removing dependence on obj_pool.h.

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: improve reporting of input errors
Jonathan Nieder [Tue, 28 Dec 2010 10:30:54 +0000 (04:30 -0600)]
vcs-svn: improve reporting of input errors

Catch input errors and exit early enough to print a reasonable
diagnosis based on errno.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: make buffer_copy_bytes return length read
Jonathan Nieder [Tue, 28 Dec 2010 10:26:17 +0000 (04:26 -0600)]
vcs-svn: make buffer_copy_bytes return length read

Currently buffer_copy_bytes does not report to its caller whether
it encountered an early end of file.

Add a return value representing the number of bytes read (but not
the number of bytes copied).  This way all three unusual conditions
can be distinguished: input error with buffer_ferror, output error
with ferror(outfile), early end of input by checking the return
value.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: make buffer_skip_bytes return length read
Jonathan Nieder [Mon, 11 Oct 2010 02:44:21 +0000 (21:44 -0500)]
vcs-svn: make buffer_skip_bytes return length read

Currently there is no way to detect when input ended if it ended
early during buffer_skip_bytes.  Tell the calling program how many
bytes were actually skipped for easier debugging.

Existing callers will still ignore early EOF.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: improve support for reading large files
Jonathan Nieder [Mon, 11 Oct 2010 02:46:24 +0000 (21:46 -0500)]
vcs-svn: improve support for reading large files

Move from uint32_t to off_t as the fundamental unit of length used by
the line_buffer library.  Performance would get worse if anything but
I think it's worth it for support of deltas that need to skip large
pieces (> 4 GiB).

Exception: buffer_read_string still takes a uint32_t, since it keeps
its result in an in-core obj_pool.

Callers still have to be updated to take advantage of this.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: allow input errors to be detected promptly
Jonathan Nieder [Mon, 11 Oct 2010 02:51:21 +0000 (21:51 -0500)]
vcs-svn: allow input errors to be detected promptly

The line_buffer library silently flags input errors until
buffer_deinit time; unfortunately, by that point usually errno is
invalid.  Expose the error flag so callers can check for and
report errors early for easy debugging.

some_error_prone_operation(...);
if (buffer_ferror(buf))
return error("input error: %s", strerror(errno));

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: simplify repo_modify_path and repo_copy
Jonathan Nieder [Fri, 10 Dec 2010 06:53:54 +0000 (00:53 -0600)]
vcs-svn: simplify repo_modify_path and repo_copy

Restrict the repo_tree API to functions that are actually needed.

 - decouple reading the mode and content of dirents from other
   operations.
 - remove repo_modify_path.  It is only used to read the mode from
   dirents.
 - remove the ability to use repo_read_mode on a missing path.  The
   existing code only errors out in that case, anyway.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: handle_node: use repo_read_path
Jonathan Nieder [Fri, 10 Dec 2010 10:28:06 +0000 (04:28 -0600)]
vcs-svn: handle_node: use repo_read_path

svn-fe processes each commit in two stages: first decide on the
correct content for all paths and export the relevant blobs, then
export a commit with the result.

But we can keep less state and simplify svn-fe a great deal by
exporting the commit in one step: use 'inline' blobs for each path and
remember nothing.  This way, the repo_tree structure could be
eliminated, and we would get support for incremental imports 'for
free'.

Reorganize handle_node along these lines.  This is just a code
cleanup; the changes in repo_tree and handle_revision will come later.

[db: backported to apply without text delta support]

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: introduce repo_read_path to check the content at a path
Jonathan Nieder [Sat, 20 Nov 2010 19:25:28 +0000 (13:25 -0600)]
vcs-svn: introduce repo_read_path to check the content at a path

The repo_tree structure remembers, for each path in each revision, a
mode (regular file, executable, symlink, or directory) and content
(blob mark or directory structure).  Maintaining a second copy of all
this information when it's already in the target repository is
wasteful, it does not persist between svn-fe invocations, and most
importantly, there is no convenient way to transfer it from one
machine to another.  So it would be nice to get rid of it.

As a first step, let's change the repo_tree API to match fast-import's
read commands more closely.  Currently to read the mode for a path,
one uses

repo_modify_path(path, new_mode, new_content);

which changes the mode and content as a side effect.  There is no
function to read the content at a path; add one.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agofast-import: make code "-Wpointer-arith" clean
Jonathan Nieder [Mon, 28 Feb 2011 21:16:59 +0000 (15:16 -0600)]
fast-import: make code "-Wpointer-arith" clean

The dereference() function to peel a tree-ish and find the underlying
tree expects arithmetic to (void *) to work on byte addresses.  We
should be reading the text of objects through a char * anyway.

Noticed-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agoMerge commit 'jn/svn-fe' of git://github.com/gitster/git into svn-fe
Jonathan Nieder [Sat, 26 Feb 2011 11:21:29 +0000 (05:21 -0600)]
Merge commit 'jn/svn-fe' of git://github.com/gitster/git into svn-fe

* git://github.com/gitster/git:
  vcs-svn: Allow change nodes for root of tree (/)
  vcs-svn: Implement Prop-delta handling
  vcs-svn: Sharpen parsing of property lines
  vcs-svn: Split off function for handling of individual properties
  vcs-svn: Make source easier to read on small screens
  vcs-svn: More dump format sanity checks
  vcs-svn: Reject path nodes without Node-action
  vcs-svn: Delay read of per-path properties
  vcs-svn: Combine repo_replace and repo_modify functions
  vcs-svn: Replace = Delete + Add
  vcs-svn: handle_node: Handle deletion case early
  vcs-svn: Use mark to indicate nodes with included text
  vcs-svn: Unclutter handle_node by introducing have_props var
  vcs-svn: Eliminate node_ctx.mark global
  vcs-svn: Eliminate node_ctx.srcRev global
  vcs-svn: Check for errors from open()
  vcs-svn: Allow simple v3 dumps (no deltas yet)

Conflicts:
t/t9010-svn-fe.sh
vcs-svn/svndump.c

13 years agovcs-svn: teach line_buffer about temporary files
Jonathan Nieder [Mon, 3 Jan 2011 03:10:59 +0000 (21:10 -0600)]
vcs-svn: teach line_buffer about temporary files

It can sometimes be useful to write information temporarily to file,
to read back later.  These functions allow a program to use the
line_buffer facilities when doing so.

It works like this:

 1. find a unique filename with buffer_tmpfile_init.
 2. rewind with buffer_tmpfile_rewind.  This returns a stdio
    handle for writing.
 3. when finished writing, declare so with
    buffer_tmpfile_prepare_to_read.  The return value indicates
    how many bytes were written.
 4. read whatever portion of the file is needed.
 5. if finished, remove the temporary file with buffer_deinit.
    otherwise, go back to step 2,

The svn support would use this to buffer the postimage from delta
application until the length is known and fast-import can receive
the resulting blob.

Based-on-patch-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: allow input from file descriptor
Jonathan Nieder [Mon, 3 Jan 2011 03:09:38 +0000 (21:09 -0600)]
vcs-svn: allow input from file descriptor

Based-on-patch-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: allow character-oriented input
Jonathan Nieder [Mon, 3 Jan 2011 03:06:32 +0000 (21:06 -0600)]
vcs-svn: allow character-oriented input

buffer_read_char can be used in place of buffer_read_string(1) to
avoid consuming valuable static buffer space.  The delta applier will
use this to read variable-length integers one byte at a time.

Underneath, it is fgetc, wrapped so the line_buffer library can
maintain its role as gatekeeper of input.

Later it might be worth checking if fgetc_unlocked is faster ---
most line_buffer functions are not thread-safe anyway.

Helpd-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: add binary-safe read function
Jonathan Nieder [Mon, 3 Jan 2011 03:05:46 +0000 (21:05 -0600)]
vcs-svn: add binary-safe read function

buffer_read_string works well for non line-oriented input except for
one problem: it does not tell the caller how many bytes were actually
written.  This means that unless one is very careful about checking
for errors (and eof) the calling program cannot tell the difference
between the string "foo" followed by an early end of file and the
string "foo\0bar\0baz".

So introduce a variant that reports the length, too, a thinner wrapper
around strbuf_fread.  Its result is written to a strbuf so the caller
does not need to keep track of the number of bytes read.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agot0081 (line-buffer): add buffering tests
Jonathan Nieder [Mon, 3 Jan 2011 01:07:16 +0000 (19:07 -0600)]
t0081 (line-buffer): add buffering tests

POSIX makes the behavior of read(2) from a pipe fairly clear: a read
from an empty pipe will block until there is data available and any
other read will not block, prefering to return a partial result.
Likewise, fread(3) and fgets(3) are clearly specified to act as
though implemented by calling fgetc(3) in a simple loop.  But the
buffering behavior of fgetc is less clear.

Luckily, no sane platform is going to implement fgetc by calling the
equivalent of read(2) more than once.  fgetc has to be able to
return without filling its buffer to preserve errno when errors are
encountered anyway.  So let's assume the simpler behavior (trust) but
add some tests to catch insane platforms that violate that when they
come (verify).

First check that fread can handle a 0-length read from an empty fifo.
Because open(O_RDONLY) blocks until the writing end is open, open the
writing end of the fifo in advance in a subshell.

Next try short inputs from a pipe that is not filled all the way.

Lastly (two tests) try very large inputs from a pipe that will not fit
in the relevant buffers.  The first of these tests reads a little
more than 8192 bytes, which is BUFSIZ (the size of stdio's buffers)
on this Linux machine.  The second reads a little over 64 KiB (the
pipe capacity on Linux) and is not run unless requested by setting
the GIT_REMOTE_SVN_TEST_BIG_FILES environment variable.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: tweak test-line-buffer to not assume line-oriented input
Jonathan Nieder [Mon, 3 Jan 2011 00:52:28 +0000 (18:52 -0600)]
vcs-svn: tweak test-line-buffer to not assume line-oriented input

Do not expect an implicit newline after each input record.
Use a separate command to exercise buffer_skip_bytes.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agotests: give vcs-svn/line_buffer its own test script
Jonathan Nieder [Mon, 3 Jan 2011 00:51:07 +0000 (18:51 -0600)]
tests: give vcs-svn/line_buffer its own test script

Split the line_buffer test into small pieces and move it to its
own file as preparation for adding more tests.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: make test-line-buffer input format more flexible
Jonathan Nieder [Mon, 3 Jan 2011 00:50:16 +0000 (18:50 -0600)]
vcs-svn: make test-line-buffer input format more flexible

Imitate the input format of test-obj-pool to support arbitrary
sequences of commands rather than alternating read/copy.  This should
make it easier to add tests that exercise other line_buffer functions.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: teach line_buffer to handle multiple input files
Jonathan Nieder [Mon, 11 Oct 2010 02:41:06 +0000 (21:41 -0500)]
vcs-svn: teach line_buffer to handle multiple input files

Collect the line_buffer state in a newly public line_buffer struct.
Callers can use multiple line_buffers to manage input from multiple
files at a time.

svn-fe's delta applier will use this to stream a delta from svnrdump
and the preimage it applies to from fast-import at the same time.

The tests don't take advantage of the new features, but I think that's
okay.  It is easier to find lingering examples of nonreentrant code by
searching for "static" in line_buffer.c.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: collect line_buffer data in a struct
Jonathan Nieder [Mon, 11 Oct 2010 02:39:21 +0000 (21:39 -0500)]
vcs-svn: collect line_buffer data in a struct

Prepare for the line_buffer lib to support input from multiple files,
by collecting global state in a struct that can be easily passed
around.

No API change yet.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: replace buffer_read_string memory pool with a strbuf
Jonathan Nieder [Sat, 6 Nov 2010 17:01:28 +0000 (12:01 -0500)]
vcs-svn: replace buffer_read_string memory pool with a strbuf

obj_pool is inherently global and does not use the standard growing
factor alloc_nr, which makes it feel out of place in the git codebase.
Plus it is overkill for this application: all that is needed is a
buffer that can grow between requests to accomodate larger strings.
Use a strbuf instead.

As a side effect, this improves the error handling: allocation
failures will result in a clean exit instead of segfaults.  It would
be nice to add a test case (using ulimit or failmalloc) but that can
wait for another day.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agovcs-svn: eliminate global byte_buffer
Jonathan Nieder [Mon, 11 Oct 2010 02:37:10 +0000 (21:37 -0500)]
vcs-svn: eliminate global byte_buffer

The data stored in byte_buffer[] is always either discarded or
written to stdout immediately.  No need for it to persist between
function calls.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
13 years agofast-import: add 'ls' command
David Barr [Thu, 2 Dec 2010 10:40:20 +0000 (21:40 +1100)]
fast-import: add 'ls' command

Lazy fast-import frontend authors that want to rely on the backend to
keep track of the content of the imported trees _almost_ have what
they need in the 'cat-blob' command (v1.7.4-rc0~30^2~3, 2010-11-28).
But it is not quite enough, since

 (1) cat-blob can be used to retrieve the content of files, but
     not their mode, and

 (2) using cat-blob requires the frontend to keep track of a name
     (mark number or object id) for each blob to be retrieved

Introduce an 'ls' command to complement cat-blob and take care of the
remaining needs.  The 'ls' command finds what is at a given path
within a given tree-ish (tag, commit, or tree):

'ls' SP <dataref> SP <path> LF

or in fast-import's active commit:

'ls' SP <path> LF

The response is a single line sent through the cat-blob channel,
imitating ls-tree output.  So for example:

FE> ls :1 Documentation
gfi> 040000 tree 9e6c2b599341d28a2a375f8207507e0a2a627fe9 Documentation
FE> ls 9e6c2b599341d28a2a375f8207507e0a2a627fe9 git-fast-import.txt
gfi> 100644 blob 4f92954396e3f0f97e75b6838a5635b583708870 git-fast-import.txt
FE> ls :1 RelNotes
gfi> 120000 blob b942e499449d97aeb50c73ca2bdc1c6e6d528743 RelNotes
FE> cat-blob b942e499449d97aeb50c73ca2bdc1c6e6d528743
gfi> b942e499449d97aeb50c73ca2bdc1c6e6d528743 blob 32
gfi> Documentation/RelNotes/1.7.4.txt

The most interesting parts of the reply are the first word, which is
a 6-digit octal mode (regular file, executable, symlink, directory,
or submodule), and the part from the second space to the tab, which is
a <dataref> that can be used in later cat-blob, ls, and filemodify (M)
commands to refer to the content (blob, tree, or commit) at that path.

If there is nothing there, the response is "missing some/path".

The intent is for this command to be used to read files from the
active commit, so a frontend can apply patches to them, and to copy
files and directories from previous revisions.

For example, proposed updates to svn-fe use this command in place of
its internal representation of the repository directory structure.
This simplifies the frontend a great deal and means support for
resuming an import in a separate fast-import run (i.e., incremental
import) is basically free.

Signed-off-by: David Barr <david.barr@cordelta.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Improved-by: Junio C Hamano <gitster@pobox.com>
Improved-by: Sverre Rabbelier <srabbelier@gmail.com>
13 years agoupdate-index --refresh --porcelain: add missing const
Jonathan Nieder [Tue, 22 Feb 2011 22:43:23 +0000 (22:43 +0000)]
update-index --refresh --porcelain: add missing const

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocheckout: add missing const to describe_detached_head
Jonathan Nieder [Tue, 22 Feb 2011 22:43:22 +0000 (22:43 +0000)]
checkout: add missing const to describe_detached_head

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Tue, 22 Feb 2011 06:46:09 +0000 (22:46 -0800)]
Merge branch 'maint'

* maint:
  Documentation: clarify -u<mode> option defaults

13 years agoDocumentation: clarify -u<mode> option defaults
Clemens Buchacher [Mon, 21 Feb 2011 20:05:25 +0000 (21:05 +0100)]
Documentation: clarify -u<mode> option defaults

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Tue, 22 Feb 2011 01:20:11 +0000 (17:20 -0800)]
Merge branch 'maint'

* maint:
  t/t7500-commit.sh: use test_cmp instead of test
  t/gitweb-lib.sh: Ensure that errors are shown for --debug --immediate
  gitweb/gitweb.perl: don't call S_ISREG() with undef
  gitweb/gitweb.perl: remove use of qw(...) as parentheses

13 years agot/t7500-commit.sh: use test_cmp instead of test
Ævar Arnfjörð Bjarmason [Sat, 19 Feb 2011 18:29:09 +0000 (18:29 +0000)]
t/t7500-commit.sh: use test_cmp instead of test

Change commit_msg_is() in t/t7500-commit.sh to use test_cmp instead of
the shell's test function. Now if a test fails we'll get test_cmp
output showing us what failed.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agot/gitweb-lib.sh: Ensure that errors are shown for --debug --immediate
Ævar Arnfjörð Bjarmason [Sat, 19 Feb 2011 18:16:19 +0000 (19:16 +0100)]
t/gitweb-lib.sh: Ensure that errors are shown for --debug --immediate

Because '--immediate' stops test suite after first error, therefore in
this mode

  test_debug 'cat gitweb.log'

was never ran, thus in effect negating effect of '--debug' option.
This made finidng the cause of errors in gitweb test sute difficult.

Modify the gitweb_run test subroutine to run test_debug itself in the
case of errors (and also remove "test_debug 'cat gitweb.log'" from
gitweb tests).

This makes it possible to run *gitweb tests* with --immediate ---debug
combination of options; also it makes gitweb tests to not output
spurious debug data that is not considered error.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Jakub Narębski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogitweb/gitweb.perl: don't call S_ISREG() with undef
Ævar Arnfjörð Bjarmason [Sat, 19 Feb 2011 15:27:42 +0000 (15:27 +0000)]
gitweb/gitweb.perl: don't call S_ISREG() with undef

Change S_ISREG($to_mode_oct) to S_ISREG($from_mode_oct) in the branch
that handles from modes, not to modes. This logic appears to have been
caused by copy/paste programming by Jakub Narebski in e8e41a93. It
would be better to rewrite this code not to be duplicated, but I
haven't done so.

This issue caused a failing test on perl 5.13.9, which has a warning
that turned this up:

     gitweb.perl: Use of uninitialized value in subroutine entry at /home/avar/g/git/t/../gitweb/gitweb.perl line 4415.

Which caused the Git test suite to fail on this test:

    ./t9500-gitweb-standalone-no-errors.sh             (Wstat: 256 Tests: 90 Failed: 84)
      Failed tests:  1-8, 10-36, 38-45, 47-48, 50-88
      Non-zero exit status: 1

Reported-by: perl 5.13.9
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Acked-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogitweb/gitweb.perl: remove use of qw(...) as parentheses
Ævar Arnfjörð Bjarmason [Sat, 19 Feb 2011 15:27:41 +0000 (15:27 +0000)]
gitweb/gitweb.perl: remove use of qw(...) as parentheses

Using the qw(...) construct as implicit parentheses was deprecated in
perl 5.13.5. Change the relevant code in gitweb to not use the
deprecated construct. The offending code was introduced in 3562198b by
Jakub Narebski.

The issue is that perl will now warn about this:

    $ perl -wE 'for my $i qw(a b) { say $i }'
    Use of qw(...) as parentheses is deprecated at -e line 1.
    a
    b

This caused gitweb.perl to warn on perl 5.13.5 and above, and these
tests to fail on those perl versions:

    ./t9501-gitweb-standalone-http-status.sh           (Wstat: 256 Tests: 11 Failed: 10)
      Failed tests:  2-11
      Non-zero exit status: 1
    ./t9502-gitweb-standalone-parse-output.sh          (Wstat: 256 Tests: 10 Failed: 9)
      Failed tests:  2-10
      Non-zero exit status: 1

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Acked-by: Jakub Narębski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Wed, 16 Feb 2011 22:33:22 +0000 (14:33 -0800)]
Merge branch 'maint'

* maint:
  parse_tag_buffer(): do not prefixcmp() out of range

13 years agoMerge branch 'maint-1.7.3' into maint
Junio C Hamano [Wed, 16 Feb 2011 22:33:11 +0000 (14:33 -0800)]
Merge branch 'maint-1.7.3' into maint

* maint-1.7.3:

13 years agoMerge branch 'maint-1.7.2' into maint-1.7.3
Junio C Hamano [Wed, 16 Feb 2011 22:32:59 +0000 (14:32 -0800)]
Merge branch 'maint-1.7.2' into maint-1.7.3

* maint-1.7.2:
  fast-import: introduce "feature notes" command
  fast-import: clarify documentation of "feature" command

13 years agoMerge branch 'maint-1.7.1' into maint-1.7.2
Junio C Hamano [Wed, 16 Feb 2011 22:32:54 +0000 (14:32 -0800)]
Merge branch 'maint-1.7.1' into maint-1.7.2

* maint-1.7.1:
  fast-import: introduce "feature notes" command
  fast-import: clarify documentation of "feature" command

13 years agoMerge branch 'maint-1.7.0' into maint-1.7.1
Junio C Hamano [Wed, 16 Feb 2011 22:32:47 +0000 (14:32 -0800)]
Merge branch 'maint-1.7.0' into maint-1.7.1

* maint-1.7.0:
  fast-import: introduce "feature notes" command
  fast-import: clarify documentation of "feature" command

13 years agoparse_tag_buffer(): do not prefixcmp() out of range
Nguyễn Thái Ngọc Duy [Mon, 14 Feb 2011 13:02:51 +0000 (20:02 +0700)]
parse_tag_buffer(): do not prefixcmp() out of range

There is a check (size < 64) at the beginning of the function, but
that only covers object+type lines.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Tue, 15 Feb 2011 19:03:22 +0000 (11:03 -0800)]
Merge branch 'maint'

* maint:
  pull: do not display fetch usage on --help-all
  git-tag.txt: list all modes in the description
  commit,status: describe -u likewise
  add: describe --patch like checkout, reset
  commit,merge,tag: describe -m likewise
  clone,init: describe --template using the same wording
  commit,status: describe --porcelain just like push
  commit,tag: use same wording for -F
  configure: use AC_LANG_PROGRAM consistently
  string_list_append: always set util pointer to NULL
  correct type of EMPTY_TREE_SHA1_BIN

13 years agopull: do not display fetch usage on --help-all
Michael J Gruber [Mon, 14 Feb 2011 16:48:08 +0000 (17:48 +0100)]
pull: do not display fetch usage on --help-all

Currently, "git pull --help-all" displays the fetch usage info.

Make it equivalent to "git pull -h" instead since "--help-all" is
documented in gitcli(7).

Do not try to sanitize the pull option parser (aka last hair puller).

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-tag.txt: list all modes in the description
Michael J Gruber [Tue, 15 Feb 2011 13:09:10 +0000 (14:09 +0100)]
git-tag.txt: list all modes in the description

Currently, the description sounds as if it applied always, but most of
its content is true in "create tag mode" only.

Make this clearer by listing all modes upfront.

Also, sneak in some linguistic improvements and make it clearer that
lightweight tags are "created" because "written" may be misread as
"are output".

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocommit,status: describe -u likewise
Michael J Gruber [Tue, 15 Feb 2011 13:09:09 +0000 (14:09 +0100)]
commit,status: describe -u likewise

They differ by one character only. Being exactly equal should help
translations.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoadd: describe --patch like checkout, reset
Michael J Gruber [Tue, 15 Feb 2011 13:09:08 +0000 (14:09 +0100)]
add: describe --patch like checkout, reset

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocommit,merge,tag: describe -m likewise
Michael J Gruber [Tue, 15 Feb 2011 13:09:07 +0000 (14:09 +0100)]
commit,merge,tag: describe -m likewise

This also removes the superfluous "specify" and rewords the misleading
"if any" which sounds as if omitting "-m" would omit the merge commit
message. (It means "if a merge commit is created at all".)

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoclone,init: describe --template using the same wording
Michael J Gruber [Tue, 15 Feb 2011 13:09:06 +0000 (14:09 +0100)]
clone,init: describe --template using the same wording

This also corrects a wrong description for clone.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocommit,status: describe --porcelain just like push
Michael J Gruber [Tue, 15 Feb 2011 13:09:05 +0000 (14:09 +0100)]
commit,status: describe --porcelain just like push

Push has the clearer description, so take that one for all.

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocommit,tag: use same wording for -F
Michael J Gruber [Tue, 15 Feb 2011 13:09:04 +0000 (14:09 +0100)]
commit,tag: use same wording for -F

Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoRevert "unpack_trees(): skip trees that are the same in all input"
Junio C Hamano [Tue, 15 Feb 2011 18:47:04 +0000 (10:47 -0800)]
Revert "unpack_trees(): skip trees that are the same in all input"

This reverts commit 83c90314aa27ae3768c04375d02e4f3fb12b726d, which
seems to have broken merge to report conflicts when there should be
none.

13 years agoconfigure: use AC_LANG_PROGRAM consistently
Ralf Wildenhues [Sun, 2 Jan 2011 10:24:55 +0000 (11:24 +0100)]
configure: use AC_LANG_PROGRAM consistently

Avoid warnings from Autoconf 2.68 about missing use of AC_LANG_PROGRAM
and friends.

Quoting autoconf-2.68/NEWS:

  ** The macros AC_PREPROC_IFELSE, AC_COMPILE_IFELSE, AC_LINK_IFELSE, and
     AC_RUN_IFELSE now warn if the first argument failed to use
     AC_LANG_SOURCE or AC_LANG_PROGRAM to generate the conftest file
     contents.  A new macro AC_LANG_DEFINES_PROVIDED exists if you have
     a compelling reason why you cannot use AC_LANG_SOURCE but must
     avoid the warning.

The underlying reason for that change is that AC_LANG_{SOURCE,PROGRAM}
take care to supply the previously computed set of #defines (and
include standard headers if so desired) for preprocessed languages
like C and C++.

In some cases, AC_LANG_PROGRAM is already used but not sufficiently
m4-quoted, so we just need to add another set of [quotes] to prevent
the autoconf warning from being triggered bogusly.  Quoting all
arguments (except when calling special macros that need to be expanded
before recursion) is better style, anyway.  These and more rules are
described in detail in 'info Autoconf "Programming in M4"'.

No change in the resulting config.mak.autogen after running
./configure intended.

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agostring_list_append: always set util pointer to NULL
Jeff King [Sat, 12 Feb 2011 05:18:51 +0000 (00:18 -0500)]
string_list_append: always set util pointer to NULL

It is not immediately obvious that the util field may
contain random bytes after appending an item. Especially
since the string_list_insert* functions _do_ explicitly zero
the util pointer.

This does not appear to be a bug in any current git code, as
all callers either fill in the util field immediately or
never use it. However, it is worth it to be less surprising
to new users of the string-list API who may expect it to be
intialized to NULL.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocorrect type of EMPTY_TREE_SHA1_BIN
Jonathan Nieder [Mon, 7 Feb 2011 08:17:27 +0000 (02:17 -0600)]
correct type of EMPTY_TREE_SHA1_BIN

Functions such as hashcmp that expect a binary SHA-1 value take
parameters of type "unsigned char *" to avoid accepting a textual
SHA-1 passed by mistake.  Unfortunately, this means passing the string
literal EMPTY_TREE_SHA1_BIN requires an ugly cast.  Tweak the
definition of EMPTY_TREE_SHA1_BIN to produce a value of more
convenient type.

In the future the definition might change to

extern const unsigned char empty_tree_sha1_bin[20];
#define EMPTY_TREE_SHA1_BIN empty_tree_sha1_bin

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoObey p4 views when using client spec
Ian Wienand [Sat, 12 Feb 2011 00:33:48 +0000 (16:33 -0800)]
Obey p4 views when using client spec

When using the p4 client spec, this attempts to obey the client's
output preferences.

For example, a view like

//depot/foo/branch/... //client/branch/foo/...
//depot/bar/branch/... //client/branch/bar/...

will result in a directory layout in the git tree of

branch/
branch/foo
branch/bar

p4 can do various other reordering that this change doesn't support,
but we should detect it and at least fail nicely.

Signed-off-by: Ian Wienand <ianw@vmware.com>
Acked-by: Pete Wyckoff <pw@padd.com>
Acked-by: Tor Arvid Lund <torarvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogitignore: add test-mktemp to ignore list
Ævar Arnfjörð Bjarmason [Sat, 12 Feb 2011 23:21:17 +0000 (23:21 +0000)]
gitignore: add test-mktemp to ignore list

Change the .gitignore to ignore test-mktemp which is built from
test-mktemp.c. Arnout Engelen added this in 6cf6bb3 (Improve error
messages when temporary file creation fails, 2010-12-18) but forgot
to add a corresponding entry to .gitignore.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorepo-config: add deprecation warning
René Scharfe [Sat, 12 Feb 2011 13:24:10 +0000 (14:24 +0100)]
repo-config: add deprecation warning

repo-config was deprecated in 5c66d0d4 on 2008-01-17.  Warn the
remaining users that it has been replaced by config and is going to
be removed eventually.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Sat, 12 Feb 2011 00:01:47 +0000 (16:01 -0800)]
Merge branch 'maint'

* maint:
  Git 1.7.4.1
  clone: fixup recurse_submodules option
  svn-fe: warn about experimental status

Conflicts:
contrib/examples/git-revert.sh
contrib/svn-fe/svn-fe.txt

13 years agoGit 1.7.4.1 v1.7.4.1
Junio C Hamano [Fri, 11 Feb 2011 22:39:55 +0000 (14:39 -0800)]
Git 1.7.4.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'jc/fsck-fixes' into maint
Junio C Hamano [Fri, 11 Feb 2011 22:26:10 +0000 (14:26 -0800)]
Merge branch 'jc/fsck-fixes' into maint

* jc/fsck-fixes:
  fsck: do not give up too early in fsck_dir()
  fsck: drop unused parameter from traverse_one_object()

13 years agoclone: fixup recurse_submodules option
Chris Packham [Thu, 10 Feb 2011 22:59:31 +0000 (11:59 +1300)]
clone: fixup recurse_submodules option

The recurse_submodules option was added in ccdd3da6 to bring 'git clone'
into line with 'git fetch' and future commands. The correct option should
have been "recurse-submodules".

Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agosvn-fe: warn about experimental status
Jonathan Nieder [Fri, 11 Feb 2011 10:36:44 +0000 (04:36 -0600)]
svn-fe: warn about experimental status

svn-fe is young and some coming cleanups might involve backward
incompatible UI changes.  Add some words of warning to the manual so
early adopters that are not following the project closely don't get
burned.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Thu, 10 Feb 2011 22:45:55 +0000 (14:45 -0800)]
Merge branch 'maint'

* maint:
  compat: helper for detecting unsigned overflow

13 years agocompat: helper for detecting unsigned overflow
Jonathan Nieder [Mon, 11 Oct 2010 02:59:26 +0000 (21:59 -0500)]
compat: helper for detecting unsigned overflow

The idiom (a + b < a) works fine for detecting that an unsigned
integer has overflowed, but a more explicit

unsigned_add_overflows(a, b)

might be easier to read.

Define such a macro, expanding roughly to ((a) < UINT_MAX - (b)).
Because the expansion uses each argument only once outside of sizeof()
expressions, it is safe to use with arguments that have side effects.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'tr/merge-unborn-clobber'
Junio C Hamano [Thu, 10 Feb 2011 00:41:17 +0000 (16:41 -0800)]
Merge branch 'tr/merge-unborn-clobber'

* tr/merge-unborn-clobber:
  Exhibit merge bug that clobbers index&WT

Conflicts:
t/t7607-merge-overwrite.sh

13 years agoMerge branch 'jc/unpack-trees'
Junio C Hamano [Thu, 10 Feb 2011 00:41:17 +0000 (16:41 -0800)]
Merge branch 'jc/unpack-trees'

* jc/unpack-trees:
  unpack_trees(): skip trees that are the same in all input
  unpack-trees.c: cosmetic fix

Conflicts:
unpack-trees.c

13 years agoMerge branch 'jc/fsck-fixes'
Junio C Hamano [Thu, 10 Feb 2011 00:41:17 +0000 (16:41 -0800)]
Merge branch 'jc/fsck-fixes'

* jc/fsck-fixes:
  fsck: do not give up too early in fsck_dir()
  fsck: drop unused parameter from traverse_one_object()

13 years agoMerge branch 'tr/diff-words-test'
Junio C Hamano [Thu, 10 Feb 2011 00:41:17 +0000 (16:41 -0800)]
Merge branch 'tr/diff-words-test'

* tr/diff-words-test:
  t4034 (diff --word-diff): add a minimum Perl drier test vector
  t4034 (diff --word-diff): style suggestions
  userdiff: simplify word-diff safeguard
  t4034: bulk verify builtin word regex sanity

13 years agoMerge branch 'rr/fi-import-marks-if-exists'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'rr/fi-import-marks-if-exists'

* rr/fi-import-marks-if-exists:
  fast-import: Introduce --import-marks-if-exists

13 years agoMerge branch 'jn/unpack-lstat-failure-report'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'jn/unpack-lstat-failure-report'

* jn/unpack-lstat-failure-report:
  unpack-trees: handle lstat failure for existing file
  unpack-trees: handle lstat failure for existing directory

13 years agoMerge branch 'ef/alias-via-run-command'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'ef/alias-via-run-command'

* ef/alias-via-run-command:
  alias: use run_command api to execute aliases

13 years agoMerge branch 'cb/setup'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'cb/setup'

* cb/setup:
  setup: translate symlinks in filename when using absolute paths

13 years agoMerge branch 'ae/better-template-failure-report'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'ae/better-template-failure-report'

* ae/better-template-failure-report:
  Improve error messages when temporary file creation fails

13 years agoMerge branch 'jn/cherry-pick-strategy-option'
Junio C Hamano [Thu, 10 Feb 2011 00:41:16 +0000 (16:41 -0800)]
Merge branch 'jn/cherry-pick-strategy-option'

* jn/cherry-pick-strategy-option:
  cherry-pick/revert: add support for -X/--strategy-option

13 years agoMerge branch 'maint-1.7.0' into maint
Junio C Hamano [Thu, 10 Feb 2011 00:40:12 +0000 (16:40 -0800)]
Merge branch 'maint-1.7.0' into maint

* maint-1.7.0:
  fast-import: introduce "feature notes" command
  fast-import: clarify documentation of "feature" command

Conflicts:
Documentation/git-fast-import.txt

13 years agofast-import: introduce "feature notes" command
Jonathan Nieder [Wed, 9 Feb 2011 22:43:57 +0000 (16:43 -0600)]
fast-import: introduce "feature notes" command

Here is a 'feature' command for streams to use to require support for
the notemodify (N) command.

When the 'feature' facility was introduced (v1.7.0-rc0~95^2~4,
2009-12-04), the notes import feature was old news (v1.6.6-rc0~21^2~8,
2009-10-09) and it was not obvious it deserved to be a named feature.
But now that is clear, since all major non-git fast-import backends
lack support for it.

Details: on git version with this patch applied, any "feature notes"
command in the features/options section at the beginning of a stream
will be treated as a no-op.  On fast-import implementations without
the feature (and older git versions), the command instead errors out
with a message like

This version of fast-import does not support feature notes.

So by declaring use of notes at the beginning of a stream, frontends
can avoid wasting time and other resources when the backend does not
support notes.  (This would be especially important for backends that
do not support rewinding history after a botched import.)

Improved-by: Thomas Rast <trast@student.ethz.ch>
Improved-by: Sverre Rabbelier <srabbelier@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agofast-import: clarify documentation of "feature" command
Jonathan Nieder [Sun, 28 Nov 2010 19:43:57 +0000 (13:43 -0600)]
fast-import: clarify documentation of "feature" command

The "feature" command allows streams to specify options for the import
that must not be ignored.  Logically, they are part of the stream,
even though technically most supported features are synonyms to
command-line options.

Make this more obvious by being more explicit about how the analogy
between most "feature" commands and command-line options works.  Treat
the feature (import-marks) that does not fit this analogy separately.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Acked-by: Sverre Rabbelier <srabbelier@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDocumentation/merge subtree How-To: fix typo
Uwe Kleine-König [Wed, 9 Feb 2011 09:04:43 +0000 (10:04 +0100)]
Documentation/merge subtree How-To: fix typo

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agopull: Document the "--[no-]recurse-submodules" options
Jens Lehmann [Mon, 7 Feb 2011 22:24:54 +0000 (23:24 +0100)]
pull: Document the "--[no-]recurse-submodules" options

In commits be254a0ea9 and 7dce19d374 the handling of the new fetch options
"--[no-]recurse-submodules" had been added to git-pull.sh. But they were
not documented as the pull options they now are, so let's fix that.

Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoquote.h: simplify the inclusion
Jonathan Nieder [Wed, 5 Jan 2011 00:36:34 +0000 (18:36 -0600)]
quote.h: simplify the inclusion

Attempting to include quote.h without first including strbuf.h results
in warnings:

 ./quote.h:33:33: warning: ‘struct strbuf’ declared inside parameter list
 ./quote.h:33:33: warning: its scope is only this definition or declaration, which is probably not what you want
 ./quote.h:34:34: warning: ‘struct strbuf’ declared inside parameter list
 ...

Add a toplevel declaration for struct strbuf to avoid this.

While at it, stop including system headers from quote.h.  git source
files already need to include git-compat-util.h sooner to ensure the
appropriate feature test macros are defined.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agosha1_object_info: examine cached_object store too
Nguyễn Thái Ngọc Duy [Sat, 5 Feb 2011 14:03:02 +0000 (21:03 +0700)]
sha1_object_info: examine cached_object store too

Cached object store was added in d66b37b (Add pretend_sha1_file()
interface. - 2007-02-04) as a way to temporarily inject some objects
to object store.

But only read_sha1_file() knows about this store. While it will return
an object from this store, sha1_object_info() will happily say
"object not found".

Teach sha1_object_info() about the cached store for consistency.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agosha1_file.c: move find_cached_object up so sha1_object_info can use it
Nguyễn Thái Ngọc Duy [Sat, 5 Feb 2011 14:03:01 +0000 (21:03 +0700)]
sha1_file.c: move find_cached_object up so sha1_object_info can use it

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoAdd const to parse_{commit,tag}_buffer()
Nguyễn Thái Ngọc Duy [Sat, 5 Feb 2011 10:52:20 +0000 (17:52 +0700)]
Add const to parse_{commit,tag}_buffer()

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agodiff: support --cached on unborn branches
Nguyễn Thái Ngọc Duy [Thu, 3 Feb 2011 06:23:34 +0000 (13:23 +0700)]
diff: support --cached on unborn branches

"git diff --cached" (without revision) used to mean "git diff --cached
HEAD" (i.e. the user was too lazy to type HEAD). This "correctly"
failed when there was no commit yet. But was that correctness useful?

This patch changes the definition of what particular command means.
It is a request to show what _would_ be committed without further "git
add". The internal implementation is the same "git diff --cached HEAD"
when HEAD exists, but when there is no commit yet, it compares the index
with an empty tree object to achieve the desired result.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogitweb: Mention optional Perl modules in INSTALL
Jakub Narebski [Thu, 3 Feb 2011 23:20:48 +0000 (00:20 +0100)]
gitweb: Mention optional Perl modules in INSTALL

Some optional additional Perl modules are required for some of extra
features.  Mention those in gitweb/INSTALL.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agopost-receive-email: suppress error if description file missing
Sitaram Chamarty [Thu, 3 Feb 2011 01:00:32 +0000 (06:30 +0530)]
post-receive-email: suppress error if description file missing

Signed-off-by: Sitaram Chamarty <sitaramc@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agot7407: fix line endings for mingw build
Pat Thoyts [Thu, 3 Feb 2011 15:31:44 +0000 (15:31 +0000)]
t7407: fix line endings for mingw build

Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agot4120-apply-popt: help systems with core.filemode=false
Johannes Sixt [Thu, 3 Feb 2011 15:31:43 +0000 (15:31 +0000)]
t4120-apply-popt: help systems with core.filemode=false

A test case verifies that filemode-only patches work as expected. Help
systems where "test -x" does not work by applying the test patch also to
the index, where the effects can be verified even on such systems.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agot3509: use unconstrained initial test to setup repository.
Pat Thoyts [Thu, 3 Feb 2011 15:31:42 +0000 (15:31 +0000)]
t3509: use unconstrained initial test to setup repository.

The first test did not run on msysGit due to the SYMLINKS constraint and
so subsequent tests failed because the test repository was not initialized.

Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agostart_command: flush buffers in the WIN32 code path as well
Johannes Sixt [Fri, 4 Feb 2011 08:41:58 +0000 (09:41 +0100)]
start_command: flush buffers in the WIN32 code path as well

The POSIX code path did The Right Thing already, but we have to do the same
on Windows.

This bug caused failures in t5526-fetch-submodules, where the output of
'git fetch --recurse-submodules' was in the wrong order.

Debugged-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agobundle: Use OFS_DELTA in bundle files
Shawn O. Pearce [Sun, 6 Feb 2011 00:58:53 +0000 (16:58 -0800)]
bundle: Use OFS_DELTA in bundle files

git-bundle first appeared in 2e0afafe ("Add git-bundle") in Feb 2007,
and first shipped in Git 1.5.1.

However, OFS_DELTA is an even earlier invention, coming about in
eb32d236 ("introduce delta objects with offset to base") in Sep 2006,
and first shipped in Git 1.4.4.5.

OFS_DELTA is smaller, about 3.2%-5% smaller, and is typically faster
to access than REF_DELTA because the exact location of the delta base
is available after parsing the object header.  Since all bundle aware
versions of Git are also OFS_DELTA aware, just make it the default.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>