Merge head 'cvs2git' of netz.smurf.noris.de/git/git

Work around git-http-pull breakage in git-fetch-script

Need to add a final slash. And make it verbose by default, since it's
so slow that otherwise people will think it's died.

git-fetch-script: use git-fetch-pack for local and ssh fetches.

Also, clean it up a lot.

Add "git_path()" and "head_ref()" helper functions.

"git_path()" returns a static pathname pointer into the git directory
using a printf-like format specifier.

"head_ref()" works like "for_each_ref()", except for just the HEAD.

Merge with Linus' current tree

Merge ... www.liacs.nl/~sverdool/git.git#cvs2git

cvsimport: getopt accepted a -q option (undocumented and unused).
Removed.

git-cvsimport-script: move working directory forward

If HEAD happened to point to a cvs branch, move the
working directory forward to the tip of the branch.
Additionally, if master and "origin" are equal,
move master forward to new origin first.

git-rev-list: make sure the output is sorted by recency

We didn't sort the refs by date, so if you had multiple refs, the end
result would not be properly sorted.

Make rev-list flush the stdio buffers after each rev.

We'd rather get the revisions in a slow but timely manner than
have to wait for them.

Make git-fetch-pack actually do all the unpacking etc.

It returns the result SHA1 on stdout, so you can do

remote=$(git-fetch-pack host:dir branchname)

and it will unpack the objects and "remote" will be the SHA1 name of the
branch on the other side. You can then save that off, or merge it, or
whatever.

Make git-fetch-pack and git-upload-pack negotiate needs/haves fully

Now the only piece missing is actually generating the pack-file.

Clean up output of "for_each_ref()" when GIT_DIR is "."

Remove the "./" at the head, it just looks much nicer.

git-cvsimport-script: remove unused variable

Commit first cut at "git-fetch-pack"

It's meant to be used by "git fetch" for the local and ssh case.

It doesn't actually do the fetching now, but it does discover the common
commit point.

Move ref path matching to connect.c library

It's a generic thing for matching refs from the other side.

Factor out the ssh connection stuff from send-pack.c

I want to use it for git-fetch-pack too.

git-cvsimport-script: more error handling

[PATCH] Fixup t/t5300 unit tests broken by 5f3de58ff85c49620ae2a1722d8d4d37c881a054

This patch fixes up the t/t5300 unit tests which were broken by the changes in:

Make the name of a pack-file depend on the objects packed there-in.

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

git-cvsimport-script: provide direct support for cvsps -z option

git-cvsimport-script: update cvsps cache instead of rebuilding it

Updating the cache is sufficient for most purposes.
If users really want to rebuild the cache, they can specify
the option themselves.

git-cvsimport-script: fix branch switching

Previous patch broke branch switching.

git-cvsimport-script: use private index.

git-rev-parse: support show sha1 names for pack entries

This is actually subtly wrong. If a short match is found in the object
directory, but would _also_ match another SHA1 ID in a pack (or it shows
in one pack but not another), we'll never have done the pack lookup, and
we think it's unique.

I can't find it in myself to care. You really want to use enough of a
SHA1 that there is never any ambiguity.

Make git-rev-parse support cogito-style "short hex names"

Currently only for unpacked objects, but the infrastructure
is there to do it for packed objects too.

git-cvsimport-script: leave working directory alone.

Make the name of a pack-file depend on the objects packed there-in.

This means that the .git/objects/pack directory is also rsync'able,
since the filenames created there-in are either unique or refer to the
same data.

Otherwise you might not be able to pull from a directory that is partly
packed without having to worry about missing objects due to pack-file
name clashes.

git-cvsimport-script: typo head -> heads

Add "git-prune-packed" that removes objects that exist in a pack.

This, together with "git repack" can be used to clean up unpacked
git archives.

Add "git repack" command that does an incremental pack

Add "--non-empty" flag to git-pack-objects

It skips writing the pack-file if it ends up being empty.

"git rev-list --unpacked" shows only unpacked commits

More infrastructure to do efficient incremental packs.

Add "--incremental" flag to git-pack-objects

It won't add an object that is already in a pack to the new pack.

Add "--all" flag to rev-parse that shows all refs

And make git-rev-list just silently ignore non-commit refs if we're not
asking for all objects.

Add "has_sha1_pack()" function to query whether the object is available in a pack

We'll want this for incremental packing.

Make git-fsck-cache check HEAD integrity

In particular, check that it's a symlink, and points to refs/heads/. We
depend on that these days not only for "git checkout", but also because
fsck and others only check for references in the .git/refs/
subdirectory, not things like HEAD itself.

Fix sparse warnings.

Mainly making a lot of local functions and variables be marked "static",
but there was a "zero as NULL" warning in there too.

cvsimport: Missing tests for verbosity flag.

Fix up header file dependencies and add sparse checking rules

We're pretty sparse-clean already, thanks to earlier efforts, but some
things inevitably creep in.

Fix up "for_each_ref()" to be more usable, and use it in git-fsck-cache

It needed to take the GIT_DIR information into account, something that
the original receive-pack usage just never cared about.

Fix gcc warning in send-pack.c

send_pack() was declared to return "int" (although nobody cared), but
didn't actually return anything.

Avoid gcc warnings in sha1_file.c

A couple of bogus "might be used undefined" warnings are avoided
by moving the initializations unnecessarily early.

Support :ext: access method.

Honour CVS_SERVER.

git-cvsimport-script: clean up documentation

Remove documentation of irrelevant "type" option.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Make specification of CVS module to convert optional.

If we're inside a checked out CVS repository, there is
no need to explicitly specify the module as it is
available in CVS/Repository.
Also read CVS/Root if it's available and -d is not specified.
Finally, explicitly pass root to cvsps as CVS/Root takes
precedence over CVSROOT.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Fixed a typo in Documentation/git-cvsimport-script.txt.

Generalize the "show each ref" code in receice-pack

This turns it into a generic "do xyz for each ref" library function.

[PATCH] Show more details of packfile with verify-pack -v.

This implements show_pack_info() function used in verify-pack
command when -v flag is used to obtain something like
unpack-objects used to give when it was first written.

It shows the following for each non-deltified object found in
the pack:

SHA1 type size offset

For deltified objects, it shows this instead:

SHA1 type size offset depth base_sha1

In order to get the output in the order that appear in the pack
file for debugging purposes, you can do this:

$ git-verify-pack -v packfile | sort -n -k 4,4

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] verify-pack updates.

Nico pointed out that having verify_pack.c and verify-pack.c was
confusing. Rename verify_pack.c to pack-check.c as suggested,
and enhances the verification done quite a bit.

- Built-in sha1_file unpacking knows that a base object of a
deltified object _must_ be in the same pack, and takes
advantage of that fact.

- Earlier verify-pack command only checked the SHA1 sum for the
entire pack file and did not look into its contents. It now
checks everything idx file claims to have unpacks correctly.

- It now has a hook to give more detailed information for
objects contained in the pack under -v flag.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Avoid unnecessarily inflating and interpreting delta

This teaches packed_delta_info() that it only needs to look at
the type of the base object to figure out both type and size of
a deltified object. This saves quite a many calls to inflate()
when dealing with a deep delta chain.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

cvsimport: add documentation.

cvsimport: Added option '-p': pass options to cvsps
Added option '-x' to cvsps call

cvsimport: Exit if an existing repository doesn't have the right branch.

cvs import: Strip whitespace at the end of the log entry
for compatibility with old cvs2git.

Do ref matching on the sender side rather than on receiver

This makes the receiver always send a full list of valid refs, which
will allow us to do better packs, as well as handle creation of new
refs. Eventually. Right now we just moved the matching and enabled it.

So now you can do

git-send-pack host:path branch1 branch2

to only send branches "branch1" and "branch2".

Add support for "forcing" a ref on the remote side

A "old ref" of all zeroes is considered a "don't care" ref, and allows
us to say "write the new ref regardless of what the old ref contained
(or even if it existed at all)".

This allows (if git-send-pack were to do it) creating new refs, and
fixing up old ones.

git-send-pack: actually send the object pack

This concludes this lesson. I've actually successfully sent an update
using the git-send-pack command.

Probably tons of work still to do, and nasty debugging, but it's now
actually potentially useful.

git-receive-pack: implement ref switch command handling

After unpacking the object pack successfully, we go through the list of
refs, and verify that they still contain their expected values. Then we
replace them with the new ones.

cvsimport: perform string comparison on "HEAD"

git-cvsimport-script: Removed redundant @old and @new clear.

git-cvsimport-script: Remove setting Sticky; it may cause problems

cvs import: Call git-update-cache multiple times
instead of with a too-long argument list.

Add comment on what send-pack still needs to do

Me tired.

git-receive-pack: start parsing ref update commands

We don't act on them yet, but we parse them.

Slow but steady progress on git pack receive/send

git-send-pack: start parsing local/remote reference differences

Right now it just shows which refs it picks up, and whether they are
the same or changed on the remote end. Getting there..

Turn on optimization again

It got turned off by mistake just because I had been doing debugging,
and committed the Makefile that had other changes ...

[PATCH] Remove unnecessary sort from t6001 testcase

This patch removes an unnecessary sort from the t6001 testcase.

Sorts were previously necessary when testing non --merge-order cases
because the output order wasn't entirely deterministic unless commit
date was fixed.

However, commit dates are now fixed, so the need for a sort has
disappeared. So the sort has been removed.

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix broken t6001 test case

This fix fixes a t/t6001 test case break that was hidden by a bug in the
test case infrastructure.

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Demonstrate broken t6001 test case function

Junio discovered a problem where an actual test case break was hidden
because pipelines are not handled properly by the test infrastructure in
t6001.

This patch fixes the broken infrastructure (and demonstrates the break
explicitly).

A subsequent patch in this series will fix the test case so that it
doesn't fail.

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix for git-rev-list --merge-order B ^A (A,B share common base) [rev 2]

This patch makes --merge-order produce the same list as git-rev-list
without --merge-order specified.

In particular, if the graph looks like this:

A
| B
|/
C
|
D

The both git-rev-list B ^A and git-rev-list --merge-order will produce B.

The unit tests have been changed to reflect the fact that the prune
points are now formally part of the start list that is used to perform
the --merge-order sort.

That is: git-rev-list --merge-order A ^D used to produce

= A
| C

It now produces:

^ A
| C

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Test case that demonstrates problem with --merge-order ^ processing

Added a test case that shows that --merge-order doesn't produce the
correct result in the following case.

A
|
| B
|/
C
|
D

git-rev-list --merge-order A ^B should produce just A. Instead
it produces BCD.

A subsequent patch will fix this defect.

Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Make send/receive-pack be closer to doing something interesting

Start of "git-send-pack", the local part of sending off a pack

Like git-receive-pack, this is only partway done.

Add first cut at "git-receive-pack"

It's not working yet, but it's at the point where I want to be able to
track my changes. The theory of operation is that this is the "remote"
side of a "git push". It can tell us what references the remote side
has, receives out reference update commands and a pack-file, and can
execute the unpacking command.

Remove bogus dup commit warning with --merge-order

It makes gitk unhappy, and besides, non-merge-order doesn't complain, so
why do it here..

Teach git-rev-list about non-commit objects

Now you can give git-rev-list tags, trees and blobs, and it will do the
proper reachability for them all. Knock wood.

Of course, you need the "--objects" flag to do anything but plain
commits.

Prepare git-rev-list for tracking tag objects too

We want to be able to just say "give a difference between these
objects", rather than limiting it to commits only. This isn't there
yet, but it sets things up to be a bit easier.

[PATCH] Adjust t5300 test for unpack-objects change

It now always read from standard input and rejects non-flag
arguments.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Re-instate dry-run logic in git-unpack-objects

It can no longer be as verbose, since it doesn't have a good way to
resolve deltas (now that it is purely streaming, it cannot seek around
to read the objects a delta is based on).

But it can check that the thing unpacks cleanly at least as far as pack
syntax goes - all the objects uncompress cleanly, and the pack has the
right final SHA1.

Clean up git-unpack-objects a bit

I'd like to add back the "dry-run" thing, but it turns out that to do it
well, I'd have to keep all the object data in memory (which is not
acceptable). So I'll clean it up a bit and make it do as many checks as
it can.

[PATCH] Add git-verify-pack command.

Given a list of <pack>.idx files, this command validates the
index file and the corresponding .pack file for consistency.

This patch also uses the same validation mechanism in fsck-cache
when the --full flag is used.

During normal operation, sha1_file.c verifies that a given .idx
file matches the .pack file by comparing the SHA1 checksum
stored in .idx file and .pack file as a minimum sanity check.
We may further want to check the pack signature and version when
we map the pack, but that would be a separate patch.

Earlier, errors to map a pack file was not flagged fatal but led
to a random fatal error later. This version explicitly die()s
when such an error is detected.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fixlets on top of Nico's clean-up.

If we prefer 0 as maxsize for diff_delta() to say "unlimited", let's be
consistent about it.

This patch also fixes type mismatch in a call to get_delta_hdr_size()
from packed_delta_info().

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] assorted delta code cleanup

This is a wrap-up patch including all the cleanups I've done to the
delta code and its usage. The most important change is the
factorization of the delta header handling code.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Fix packed_delta_info() that was broken by the delta header packing change

Pointed out by Junio.

Make git pack files use little-endian size encoding

This makes it match the new delta encoding, and admittedly makes the
code easier to follow.

This also updates the PACK file version to 2, since this (and the delta
encoding change in the previous commit) are incompatible with the old
format.

[PATCH] denser delta header encoding

Since the delta data format is not tied to any actual git object
anymore, now is the time to add a small improvement to the delta data
header as it is been done for packed object header. This patch allows
for reducing the delta header of about 2 bytes and makes for simpler
code.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Emit base objects of a delta chain when the delta is output.

Deltas are useless by themselves and when you use them you need to get
to their base objects. A base object should inherit recency from the
most recent deltified object that is based on it and that is what this
patch teaches git-pack-objects.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

git-unpack-objects: re-write to read from stdin

It gets a bit more complicated to unpack in a streaming environment, but
here it is. The rewrite is actually a lot cleaner in other ways, it's
just a bit more subtle.

[PATCH] Use enhanced diff_delta() in the similarity estimator.

The diff_delta() interface was extended to reject generating too big a
delta while we were working on the packed GIT archive format.

Take advantage of that when generating delta in the similarity estimator
used in diffcore-rename.c

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix unpack-objects for header length information.

Standalone unpack-objects command was not adjusted for header length
encoding change when dealing with deltified entry. This fixes it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge fighting fsck-cache updates from Junio

It's just easier to let git help out with the merge than it is to try to
fix up the diffs.

[PATCH] Update fsck-cache (take 2)

The fsck-cache complains if objects referred to by files in .git/refs/
or objects stored in files under .git/objects/??/ are not found as
stand-alone SHA1 files (i.e. found in alternate object pools
GIT_ALTERNATE_OBJECT_DIRECTORIES or packed archives stored under
.git/objects/pack).

Although this is a good semantics to maintain consistency of a single
.git/objects directory as a self contained set of objects, it sometimes
is useful to consider it is OK as long as these "outside" objects are
available.

This commit introduces a new flag, --standalone, to git-fsck-cache.
When it is not specified, connectivity checks and .git/refs pointer
checks are taught that it is OK when expected objects do not exist under
.git/objects/?? hierarchy but are available from an packed archive or in
an alternate object pool.

Another new flag, --full, makes git-fsck-cache to check not only the
current GIT_OBJECT_DIRECTORY but also objects found in alternate object
pools and packed GIT archives.a

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Expose packed_git and alt_odb.

The commands git-fsck-cache and probably git-*-pull needs to have a way
to enumerate objects contained in packed GIT archives and alternate
object pools. This commit exposes the data structure used to keep track
of them from sha1_file.c, and adds a couple of accessor interface
functions for use by the enhanced git-fsck-cache command.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Bugfix: initialize pack_base to NULL.

This was causing random segfaults, because use_packed_git() got
confused by random garbage there.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Adjust fsck-cache to packed GIT and alternate object pool.

The fsck-cache complains if objects referred to by files in .git/refs/
or objects stored in files under .git/objects/??/ are not found as
stand-alone SHA1 files (i.e. found in alternate object pools
GIT_ALTERNATE_OBJECT_DIRECTORIES or packed archives stored under
.git/objects/pack).

Although this is a good semantics to maintain consistency of a single
.git/objects directory as a self contained set of objects, it sometimes
is useful to consider it is OK as long as these "outside" objects are
available.

This commit introduces a new flag, --standalone, to git-fsck-cache.
When it is not specified, connectivity checks and .git/refs pointer
checks are taught that it is OK when expected objects do not exist under
.git/objects/?? hierarchy but are available from an packed archive or in
an alternate object pool.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Change pack file format. Hopefully for the last time.

This also adds a header with a signature, version info, and the number
of objects to the pack file. It also encodes the file length and type
more efficiently.

More doc

Document the new migration tool