delta/git.git, branch rs/ref-transaction-multi

### rs/ref-transaction-multi

2014-09-03T17:17:05+00:00

refs.c: make the *_packed_refs functions static

2014-09-03T17:16:23+00:00

We no longer need to expose the lock/add/commit/rollback functions
for packed refs anymore so make them static and remove them from the
public api.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

refs.c: make repack_without_refs static

2014-09-03T17:16:23+00:00

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

remote.c: use a transaction for deleting refs

2014-09-03T17:16:23+00:00

Transactions now use packed refs when deleting multiple refs so there is no
need to do it manually from remote.c any more.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

refs.c: write updates to packed refs when a transaction has more than one ref

2014-09-03T17:16:23+00:00

When we are updating more than one single ref, i.e. not a commit, then
write the updated refs directly to the packed refs file instead of writing
them as loose refs.

Change clone to use a transaction instead of using the packed refs API.
This changes the behavior of clone slightly. Previously clone would always
clone all refs into a packed refs file. With this change clone will only
clone into packed refs iff there are two or more refs being cloned.
If the repository we are cloning from only contains exactly one single ref
then clone will now store this as a loose ref. The benefit here is that
we no longer need to export a bunch of API functions to clone to access
packed refs directly. Clone can now just use a normal transaction and all
the packed refs goodness will happen automatically.

Update the t5516 test to cope with the fact that clone now only uses
packed refs if there are more than one ref being cloned.

We still use loose refs for single ref transactions, such as are used
by 'git commit' and friends. The reason for this is that if you have very
many refs then having to re-write the whole packed refs file for every
common operation like commit would have a performance impact.
That said, with these changes it should now be fairly straightforward to
add support to optionally start using packed refs for ALL updates
which could solve existing issues with name clashes in case insensitive
filesystems.

This change also means that multi-ref updates will now appear as a single
atomic change to any external observers instead of a sequence of discreete
changes.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

refs.c: move reflog updates into its own function

2014-09-03T17:16:23+00:00

write_ref_sha1 tries to update the reflog while updating the ref.
Move these reflog changes out into its own function so that we can do the
same thing if we write a sha1 ref differently, for example by writing a ref
to the packed refs file instead.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

### rs/ref-transaction-rename

2014-09-03T17:16:18+00:00

refs.c: rollback the lockfile before we die() in repack_without_refs

2014-09-03T17:16:06+00:00

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

refs.c: update rename_ref to use a transaction

2014-09-03T17:16:06+00:00

Change refs.c to use a single transaction to copy/rename both the refs and
its reflog. Since we are no longer using rename() to move the reflog file
we no longer need to disallow rename_ref for refs with a symlink for its
reflog so we can remove that test from the testsuite.

Change the function to return 1 on failure instead of either -1 or 1.

These changes make the rename_ref operation atomic. This also eliminates the
need to use rename() to shift the reflog around via a temporary filename.
As an extension to this, since we no longer use rename() on the reflog file,
we can now safely perform renames even if the reflog is a symbolic link
and thus can remove the check and fail for that condition.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano

refs.c: use packed refs when deleting refs during a transaction

2014-09-03T17:16:06+00:00

Make the deletion of refs during a transaction more atomic.
Start by first copying all loose refs we will be deleting to the packed
refs file and then commit the packed refs file. Then re-lock the packed refs
file to stop anyone else from modifying these refs and keep it locked until
we are finished.
Since all refs we are about to delete are now safely held in the packed refs
file we can proceed to immediately unlink any corresponding loose refs
and still be fully rollback-able.

The exception is for refs that can not be resolved. Those refs are never
added to the packed refs and will just be un-rollback-ably deleted during
commit.

By deleting all the loose refs at the start of the transaction we make make
it possible to both delete one ref and then re-create a different ref in
the same transaction even if the two refs would normally conflict.

Example: rename m->m/m

In that example we want to delete the file 'm' so that we make room so
that we can create a directory with the same name in order to lock and
write to the ref m/m and its lock-file m/m.lock.

If there is a failure during the commit phase we can rollback without losing
any refs since we have so far only deleted loose refs that that are
guaranteed to also have a corresponding entry in the packed refs file.
Once we have finished all updates for refs and their reflogs we can repack
the packed refs file and remove the to-be-deleted refs from the packed refs,
at which point all the deleted refs will disappear in one atomic rename
operation.

This also means that for an outside observer, deletion of multiple refs
in a single transaction will look atomic instead of one ref being deleted
at a time.

In order to do all this we need to change the semantics for the
repack_without_refs function so that we can lock the packed refs file,
do other stuff, and later be able to call repack_without_refs with the
lock already taken.
This means we need some additional changes in remote.c to reflect the
changes to the repack_without_refs semantics.

Signed-off-by: Ronnie Sahlberg 
Signed-off-by: Junio C Hamano