When you have grafts that pretend that a given commit has different
parents than the ones recorded in the commit object, it is dangerous
to let 'git repack' remove those hidden parents, as you can easily
remove the graft and end up with a broken repository.
So let's play it safe and keep those parent objects and everything
that is reachable by them, in addition to the grafted parents.
As this behavior can only be triggered by git pack-objects, and as that
command handles duplicate parents gracefully, we do not bother to cull
duplicated parents that may result by using both true and grafted
parents.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change calls to die(..., strerror(errno)) to use the new die_errno().
In the process, also make slight style adjustments: at least state
_something_ about the function that failed (instead of just printing
the pathname), and put paths in single quotes.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Shifting 'unsigned char' or 'unsigned short' left can result in sign
extension errors, since the C integer promotion rules means that the
unsigned char/short will get implicitly promoted to a signed 'int' due to
the shift (or due to other operations).
This normally doesn't matter, but if you shift things up sufficiently, it
will now set the sign bit in 'int', and a subsequent cast to a bigger type
(eg 'long' or 'unsigned long') will now sign-extend the value despite the
original expression being unsigned.
One example of this would be something like
unsigned long size;
unsigned char c;
size += c << 24;
where despite all the variables being unsigned, 'c << 24' ends up being a
signed entity, and will get sign-extended when then doing the addition in
an 'unsigned long' type.
Since git uses 'unsigned char' pointers extensively, we actually have this
bug in a couple of places.
I may have missed some, but this is the result of looking at
git grep '[^0-9 ][ ]*<<[ ][a-z]' -- '*.c' '*.h'
git grep '<<[ ]*24'
which catches at least the common byte cases (shifting variables by a
variable amount, and shifting by 24 bits).
I also grepped for just 'unsigned char' variables in general, and
converted the ones that most obviously ended up getting implicitly cast
immediately anyway (eg hash_name(), encode_85()).
In addition to just avoiding 'unsigned char', this patch also tries to use
a common idiom for the delta header size thing. We had three different
variations on it: "& 0x7fUL" in one place (getting the sign extension
right), and "& ~0x80" and "& 0x7f" in two other places (not getting it
right). Apart from making them all just avoid using "unsigned char" at
all, I also unified them to then use a simple "& 0x7f".
I considered making a sparse extension which warns about doing implicit
casts from unsigned types to signed types, but it gets rather complex very
quickly, so this is just a hack.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In particular, pushing the "path_name()" call _into_ the show() function
would seem to allow
- more clarity into who "owns" the name (ie now when we free the name in
the show_object callback, it's because we generated it ourselves by
calling path_name())
- not calling path_name() at all, either because we don't care about the
name in the first place, or because we are actually happy walking the
linked list of "struct name_path *" and the last component.
Now, I didn't do that latter optimization, because it would require some
more coding, but especially looking at "builtin-pack-objects.c", we really
don't even want the whole pathname, we really would be better off with the
list of path components.
Why? We use that name for two things:
- add_preferred_base_object(), which actually _wants_ to traverse the
path, and now does it by looking for '/' characters!
- for 'name_hash()', which only cares about the last 16 characters of a
name, so again, generating the full name seems to be just unnecessary
work.
Anyway, so I didn't look any closer at those things, but it did convince
me that the "show_object()" calling convention was crazy, and we're
actually better off doing _less_ in list-objects.c, and giving people
access to the internal data structures so that they can decide whether
they want to generate a path-name or not.
This patch does that, and then for people who did use the name (even if
they might do something more clever in the future), it just does the
straightforward "name = path_name(path, component); .. free(name);" thing.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Here's a less trivial thing, and slightly more dubious one.
I was looking at that "struct object_array objects", and wondering why we
do that. I have honestly totally forgotten. Why not just call the "show()"
function as we encounter the objects? Rather than add the objects to the
object_array, and then at the very end going through the array and doing a
'show' on all, just do things more incrementally.
Now, there are possible downsides to this:
- the "buffer using object_array" _can_ in theory result in at least
better I-cache usage (two tight loops rather than one more spread out
one). I don't think this is a real issue, but in theory..
- this _does_ change the order of the objects printed. Instead of doing a
"process_tree(revs, commit->tree, &objects, NULL, "");" in the loop
over the commits (which puts all the root trees _first_ in the object
list, this patch just adds them to the list of pending objects, and
then we'll traverse them in that order (and thus show each root tree
object together with the objects we discover under it)
I _think_ the new ordering actually makes more sense, but the object
ordering is actually a subtle thing when it comes to packing
efficiency, so any change in order is going to have implications for
packing. Good or bad, I dunno.
- There may be some reason why we did it that odd way with the object
array, that I have simply forgotten.
Anyway, now that we don't buffer up the objects before showing them
that may actually result in lower memory usage during that whole
traverse_commit_list() phase.
This is seriously not very deeply tested. It makes sense to me, it seems
to pass all the tests, it looks ok, but...
Does anybody remember why we did that "object_array" thing? It used to be
an "object_list" a long long time ago, but got changed into the array due
to better memory usage patterns (those linked lists of obejcts are
horrible from a memory allocation standpoint). But I wonder why we didn't
do this back then. Maybe there's a reason for it.
Or maybe there _used_ to be a reason, and no longer is.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On Wed, 8 Apr 2009, Björn Steinbrink wrote:
>
> The name of the processed object was duplicated for passing it to
> add_object(), but that already calls path_name, which allocates a new
> string anyway. So the memory allocated by the xstrdup calls just went
> nowhere, leaking memory.
Ack, ack.
There's another easy 5% or so for the built-in object walker: once we've
created the hash from the name, the name isn't interesting any more, and
so something trivial like this can help a bit.
Does it matter? Probably not on its own. But a few more memory saving
tricks and it might all make a difference.
Linus
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the case of a small repository, pack-objects is smart enough to not
start more threads than necessary. However, the output to the user always
reports the value of the pack.threads configuration and not the real
number of threads to be used.
Signed-off-by: Dan McGee <dpmcgee@gmail.com>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The goal of this patch is to get rid of the "static struct rev_info
revs" static variable in "builtin-rev-list.c".
To do that, we need to pass the revs to the "show_commit" function
in "builtin-rev-list.c" and this in turn means that the
"traverse_commit_list" function in "list-objects.c" must be passed
functions pointers to functions with 2 parameters instead of one.
So we have to change all the callers and all the functions passed
to "traverse_commit_list".
Anyway this makes the code more clean and more generic, so it
should be a good thing in the long run.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On a 32-bit system, the maximum possible size for an object is less than
4GB, while 64-bit systems may cope with larger objects. Due to this
limitation, variables holding object sizes are using an unsigned long
type (32 bits on 32-bit systems, or 64 bits on 64-bit systems).
When large objects are encountered, and/or people play with large delta
depth values, it is possible for the maximum allowed delta size
computation to overflow, especially on a 32-bit system. When this
occurs, surviving result bits may represent a value much smaller than
what it is supposed to be, or even zero. This prevents some objects
from being deltified although they do get deltified when a smaller depth
limit is used. Fix this by always performing a 64-bit multiplication.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If pack-objects is called with the --unpack-unreachable option then it
will unpack (i.e. loosen) all unreferenced objects from local not-kept
packs, including those that also exist in packs residing in an alternate
object database or a locally kept pack. The only user of this option is
git-repack.
In this case, repack will follow the call to pack-objects with a call to
prune-packed, which will delete these newly loosened objects, making the
act of loosening a waste of time. The unnecessary loosening can be
avoided by checking whether an object exists in a non-local pack or a
locally kept pack before loosening it.
This fixes the 'local packed unreachable obs that exist in alternate ODB
are not loosened' test in t7700.
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This option to pack-objects/rev-list was created to improve the -A and -a
options of repack. It was found to be lacking in that it did not provide
the ability to differentiate between local and non-local kept packs, and
found to be unnecessary since objects residing in local kept packs can be
filtered out by the --honor-pack-keep option.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These two features were invented for use by repack when repack will delete
the local packs that have been made redundant. The packs accessible
through alternates are not deleted by repack, so the objects contained in
them are still accessible after the local packs are deleted. They do not
need to be repacked into the new pack or loosened. For the case of
loosening they would immediately be deleted by the subsequent prune-packed
that is called by repack anyway.
This fixes the test
'packed unreachable obs in alternate ODB are not loosened' in t7700.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now is_kept_pack() is just a member lookup into a structure, we can write
it as such.
Also rewrite the sole caller of has_sha1_kept_pack() to switch on the
criteria the callee uses (namely, revs->kept_pack_only) between calling
has_sha1_kept_pack() and has_sha1_pack(), so that these two callees do not
have to take a pointer to struct rev_info as an argument.
This removes the header file dependency issue temporarily introduced by
the earlier commit, so we revert changes associated to that as well.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This removes --unpacked=<packfile> parameter from the revision parser, and
rewrites its use in git-repack to pass a single --kept-pack-only option
instead.
The new --kept-pack-only option means just that. When this option is
given, is_kept_pack() that used to say "not on the --unpacked=<packfile>
list" now says "the packfile has corresponding .keep file".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This refactors three loops that check if a given packfile is on the
ignore_packed list into a function is_kept_pack(). The function returns
false for a pack on the list, and true for a pack not on the list, because
this list is solely used by "git repack" to pass list of packfiles that do
not have corresponding .keep files, i.e. a packfile not on the list is
"kept".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a repository created with git older than f49fb35 (git-init-db: create
"pack" subdirectory under objects, 2005-06-27), objects/pack/ directory is
not created upon initialization. It was Ok because subdirectories are
created as needed inside directories init-db creates, and back then,
packfiles were recent invention.
After the said commit, new codepaths started relying on the presense of
objects/pack/ directory in the repository. This was exacerbated with
8b4eb6b (Do not perform cross-directory renames when creating packs,
2008-09-22) that moved the location temporary pack files are created from
objects/ directory to objects/pack/ directory, because moving temporary to
the final location was done carefully with lazy leading directory creation.
Many packfile related operations in such an old repository can fail
mysteriously because of this.
This commit introduces two helper functions to make things work better.
- odb_mkstemp() is a specialized version of mkstemp() to refactor the
code and teach it to create leading directories as needed;
- odb_pack_keep() refactors the code to create a ".keep" file while
create leading directories as needed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
R. Tyler Ballance reported a mysterious transient repository corruption;
after much digging, it turns out that we were not catching and reporting
memory allocation errors from some calls we make to zlib.
This one _just_ wraps things; it doesn't do the "retry on low memory
error" part, at least not yet. It is an independent issue from the
reporting. Some of the errors are expected and passed back to the caller,
but we die when zlib reports it failed to allocate memory for now.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If there are few objects to deltify, they might be split amongst threads
so that there is simply no other objects left to delta against within
the same thread. Let's use the same 2*window treshold as used for the
final load balancing to allow extra threads to be created.
This fixes the benign t5300 test failure.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Tested-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
... and display the actual number of threads used when locally
repacking. A remote server still won't tell you how many threads it
uses during a fetch though.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Especially on Windows where an opened file cannot be replaced, make
sure pack-objects always close packs it is about to replace. Even on
non Windows systems, this could save potential bad results if ever
objects were to be read from the new pack file using offset from the old
index.
This should fix t5303 on Windows.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Tested-by: Johannes Sixt <j6t@kdbg.org> (MinGW)
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the limit was sufficiently low, having a single object written
could bust the limit (by design), but caused the remaining allowed
size to go negative for subsequent objects, which for an unsigned
variable is a rather huge limit.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With this patch, --local means pack only local objects that are not already
packed.
Additionally, this fixes t7700 testing whether loose objects in an alternate
object database are repacked.
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This adds a new option to pack-objects which will cause it to ignore an
object which appears in a local pack which has a .keep file, even if it
was specified for packing.
This option will be used by the porcelain repack.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the pack data to be reused is found to be bad, let's fall back to
full object access through the generic path which has its own strategies
to find alternate object sources in that case. This allows for "fixing"
a corrupted pack simply by copying either another pack containing the
object(s) found to be bad, or the loose object itself, into the object
store and launch a repack without the need for -f.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It currently calls die() whenever given offset is not found thinking
that such thing should never happen. But this offset may come from a
corrupted pack whych _could_ happen and not be found. Callers should
deal with this possibility gracefully instead.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The check_object() function tries to get away with the least amount of
pack access possible when it already has partial information on given
object rather than calling the more costly packed_object_info().
When things don't look right, it should just give up and fall back to
packed_object_info() directly instead of die()'ing.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is possible to have pack corruption in the object header. Currently
unpack_object_header() simply die() on them instead of letting the caller
deal with that gracefully.
So let's have unpack_object_header() return an error instead, and find
a better name for unpack_object_header_gently() in that context. All
callers of unpack_object_header() are ready for it.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In one case, it was possible to have a bad offset equal to 0 effectively
pointing a delta onto itself and crashing git after too many recursions.
In the other cases, a negative offset could result due to off_t being
signed. Catch those.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Abstract
--------
With index v2 we have a per object CRC to allow quick and safe reuse of
pack data when repacking. This, however, doesn't currently prevent a
stealth corruption from being propagated into a new pack when _not_
reusing pack data as demonstrated by the modification to t5302 included
here.
The Context
-----------
The Git database is all checksummed with SHA1 hashes. Any kind of
corruption can be confirmed by verifying this per object hash against
corresponding data. However this can be costly to perform systematically
and therefore this check is often not performed at run time when
accessing the object database.
First, the loose object format is entirely compressed with zlib which
already provide a CRC verification of its own when inflating data. Any
disk corruption would be caught already in this case.
Then, packed objects are also compressed with zlib but only for their
actual payload. The object headers and delta base references are not
deflated for obvious performance reasons, however this leave them
vulnerable to potentially undetected disk corruptions. Object types
are often validated against the expected type when they're requested,
and deflated size must always match the size recorded in the object header,
so those cases are pretty much covered as well.
Where corruptions could go unnoticed is in the delta base reference.
Of course, in the OBJ_REF_DELTA case, the odds for a SHA1 reference to
get corrupted so it actually matches the SHA1 of another object with the
same size (the delta header stores the expected size of the base object
to apply against) are virtually zero. In the OBJ_OFS_DELTA case, the
reference is a pack offset which would have to match the start boundary
of a different base object but still with the same size, and although this
is relatively much more "probable" than in the OBJ_REF_DELTA case, the
probability is also about zero in absolute terms. Still, the possibility
exists as demonstrated in t5302 and is certainly greater than a SHA1
collision, especially in the OBJ_OFS_DELTA case which is now the default
when repacking.
Again, repacking by reusing existing pack data is OK since the per object
CRC provided by index v2 guards against any such corruptions. What t5302
failed to test is a full repack in such case.
The Solution
------------
As unlikely as this kind of stealth corruption can be in practice, it
certainly isn't acceptable to propagate it into a freshly created pack.
But, because this is so unlikely, we don't want to pay the run time cost
associated with extra validation checks all the time either. Furthermore,
consequences of such corruption in anything but repacking should be rather
visible, and even if it could be quite unpleasant, it still has far less
severe consequences than actively creating bad packs.
So the best compromize is to check packed object CRC when unpacking
objects, and only during the compression/writing phase of a repack, and
only when not streaming the result. The cost of this is minimal (less
than 1% CPU time), and visible only with a full repack.
Someone with a stats background could provide an objective evaluation of
this, but I suspect that it's bad RAM that has more potential for data
corruptions at this point, even in those cases where this extra check
is not performed. Still, it is best to prevent a known hole for
corruption when recreating object data into a new pack.
What about the streamed pack case? Well, any client receiving a pack
must always consider that pack as untrusty and perform full validation
anyway, hence no such stealth corruption could be propagated to remote
repositoryes already. It is therefore worthless doing local validation
in that case.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the main loop of find_deltas, we do:
struct object_entry *entry = *list++;
...
if (!*list_size)
...
break
Because we look at and increment *list _before_ the check of
list_size, in the very last iteration of the loop we will
look at uninitialized data, and increment the pointer beyond
one past the end of the allocated space. Since we don't
actually do anything with the data until after the check,
this is not a problem in practice.
But since it technically violates the C standard, and
because it provokes a spurious valgrind warning, let's just
move the initialization of entry to a safe place.
This fixes valgrind errors in t5300, t5301, t5302, t303, and
t9400.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many call sites immediately initialize allocated memory with zero after
calling xmalloc. A single call to xcalloc can replace this two-call
sequence.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
A comment on top of create_tmpfile() describes caveats ('can have
problems on various systems (FAT, NFS, Coda)') that should apply
in this situation as well. This in the end did not end up solving
any of my personal problems, but it might be a useful cleanup patch
nevertheless.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
User notifications are presented as 'git cmd', and code comments
are presented as '"cmd"' or 'git's cmd', rather than 'git-cmd'.
Signed-off-by: Heikki Orsila <heikki.orsila@iki.fi>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This improves commit 6d6f9cddbe a bit by simply not including missing
bases in the list of objects to process at all.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When limiting the pack size, a new header has to be written to the
pack and a new SHA1 computed. Make sure that the SHA1 of what is being
read back matches the SHA1 of what was written.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, this function has the potential to read corrupted pack data
from disk and give it a valid SHA1 checksum. Let's add the ability to
validate SHA1 checksum of existing data along the way, including before
and after any arbitrary point in the pack.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function returns 0 when the current object couldn't be written
due to the pack size limit, otherwise the current offset in the pack.
There is a problem with this approach however, since current object
could be a delta and its delta base might just have been written in
the same write_one() call, but those successfully written objects are
not accounted in the offset variable tracked by the caller. Currently
this is not an issue but a subsequent patch will need this.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The index-pack command, when processing a thin pack, fixed up the pack
after-the-fact. It forgets to fsync the result, because it only did that
in one path rather in all cases of fixup.
This moves the fsync_or_die() to the fix-up routine itself, rather than
doing it in one of the callers, so that all cases are covered.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we are building a thin pack and one of the base objects we would
consider for deltification is missing its OK, the other side already
has that base object. We may be able to get a delta from another
object, or we can simply send the new object whole (no delta).
This change allows a shallow clone to store only the objects which
are unique to it, as well as the boundary commit and its trees, but
avoids storing the boundary blobs. This special form of a shallow
clone is able to represent just the difference between two trees.
Pack objects change suggested by Nicolas Pitre.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When printing valuds of type uint32_t, we should use PRIu32, and should
not assume that it is unsigned int. On 32-bit platforms, it could be
defined as unsigned long. The same caution applies to ntohl().
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To do so, check_pack_crc() moved from builtin-pack-objects.c to
pack-check.c where it is more logical to share.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This makes life much easier for next patch, as well as being more efficient
when the revindex is actually not used.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the pack-files are now always created stably on disk, there is no
need to sync() before pruning lose objects or old stale pack-files.
[jc: with Nico's clean-up]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This means that we can depend on packs always being stable on disk,
simplifying a lot of the object serialization worries. And unlike loose
objects, serializing pack creation IO isn't going to be a performance
killer.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git_config() only had a function parameter, but no callback data
parameter. This assumes that all callback functions only modify
global variables.
With this patch, every callback gets a void * parameter, and it is hoped
that this will help the libification effort.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit ccc1297226 changed the behavior
of 'git repack -A' so unreachable objects are stored as loose objects.
However it did so in a naive and inn efficient way by making packs
about to be deleted inaccessible and feeding their content through
'git unpack-objects'. While this works, there are major flaws with
this approach:
- It is unacceptably sloooooooooooooow.
In the Linux kernel repository with no actual unreachable objects,
doing 'git repack -A -d' before:
real 2m33.220s
user 2m21.675s
sys 0m3.510s
And with this change:
real 0m36.849s
user 0m24.365s
sys 0m1.950s
For reference, here's the timing for 'git repack -a -d':
real 0m35.816s
user 0m22.571s
sys 0m2.011s
This is explained by the fact that 'git unpack-objects' was used to
unpack _every_ objects even if (almost) 100% of them were thrown away.
- There is a black out period.
Between the removal of the .idx file for the redundant pack and the
completion of its unpacking, the unreachable objects become completely
unaccessible. This is not a big issue as we're talking about unreachable
objects, but some consistency is always good.
- There is no way to easily set a sensible mtime for the newly created
unreachable loose objects.
So, while having a command called "pack-objects" to perform object
unpacking looks really odd, this is probably the best compromize to be
able to solve the above issues in an efficient way.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'depth' variable doesn't reflect the actual maximum depth used
when other objects already depend on the current one.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the delta data is cached in memory until it is written to a pack
file on disk, it is best to compress it right away in find_deltas() for
the following reasons:
- we have to compress that data anyway;
- this allows for caching more deltas with the same cache size limit;
- compression is potentially threaded.
This last point is especially relevant for SMP run time. For example,
repacking the Linux repo on a quad core processor using 4 threads with
all default settings produce the following results before this change:
real 2m27.929s
user 4m36.492s
sys 0m3.091s
And with this change applied:
real 2m13.787s
user 4m37.486s
sys 0m3.159s
So the actual execution time stayed more or less the same but the
wall clock time is shorter.
This is however not a good thing to do when generating a pack for
network transmission. In that case, the network is most likely to
throttle the data throughput, so it is best to make find_deltas()
faster in order to start writing data ASAP since we can afford
spending more time between writes to compress the data
at that point.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Johannes Schindelin