Since v1.6.0.2~13^2~ the completion of a thin pack uses sha1write() for
its ability to compute a SHA1 on the written data. This also provides
data buffering which, along with commit 92392b4a45, will confuse pread()
whenever an appended object is 1) freed due to memory pressure because
of the depth-first delta processing, and 2) needed again because it has
many delta children, and 3) its data is still buffered by sha1write().
Let's fix the issue by simply forcing cached data out when such an
object is written so it can be pread()'d at leisure.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
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>
The provided name argument is always constant and valid in every
caller's context, so no need to have an array of PATH_MAX chars to copy
it into when a simple pointer will do. Unfortunately that means getting
rid of wascally wabbits too.
The 'error' field is also unused.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The throughput display needs a delay period before accounting and
displaying anything. Yet it might be called after some amount of data
has already been transferred. The display of total data is therefore
accounted late and therefore smaller than the reality.
Let's call display_throughput() with an absolute amount of transferred
data instead of a relative number, and let the throughput code find the
relative amount of data by itself as needed. This way the displayed
total is always exact.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This one triggers only when git-pack-objects is called with
--all-progress and --stdout which is the combination used by
git-push.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add config variables pack.compression and core.loosecompression ,
and switch --compression=level to pack-objects.
Loose objects will be compressed using core.loosecompression if set,
else core.compression if set, else Z_BEST_SPEED.
Packed objects will be compressed using --compression=level if seen,
else pack.compression if set, else core.compression if set,
else Z_DEFAULT_COMPRESSION. This is the "pack compression level".
Loose objects added to a pack undeltified will be recompressed
to the pack compression level if it is unequal to the current
loose compression level by the preceding rules, or if the loose
object was written while core.legacyheaders = true. Newly
deltified loose objects are always compressed to the current
pack compression level.
Previously packed objects added to a pack are recompressed
to the current pack compression level exactly when their
deltification status changes, since the previous pack data
cannot be reused.
In either case, the --no-reuse-object switch from the first
patch below will always force recompression to the current pack
compression level, instead of assuming the pack compression level
hasn't changed and pack data can be reused when possible.
This applies on top of the following patches from Nicolas Pitre:
[PATCH] allow for undeltified objects not to be reused
[PATCH] make "repack -f" imply "pack-objects --no-reuse-object"
Signed-off-by: Dana L. How <danahow@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
The most important optimization for performance when repacking is the
ability to reuse data from a previous pack as is and bypass any delta
or even SHA1 computation by simply copying the raw data from one pack
to another directly.
The problem with this is that any data corruption within a copied object
would go unnoticed and the new (repacked) pack would be self-consistent
with its own checksum despite containing a corrupted object. This is a
real issue that already happened at least once in the past.
In some attempt to prevent this, we validate the copied data by inflating
it and making sure no error is signaled by zlib. But this is still not
perfect as a significant portion of a pack content is made of object
headers and references to delta base objects which are not deflated and
therefore not validated when repacking actually making the pack data reuse
still not as safe as it could be.
Of course a full SHA1 validation could be performed, but that implies
full data inflating and delta replaying which is extremely costly, which
cost the data reuse optimization was designed to avoid in the first place.
So the best solution to this is simply to store a CRC32 of the raw pack
data for each object in the pack index. This way any object in a pack can
be validated before being copied as is in another pack, including header
and any other non deflated data.
Why CRC32 instead of a faster checksum like Adler32? Quoting Wikipedia:
Jonathan Stone discovered in 2001 that Adler-32 has a weakness for very
short messages. He wrote "Briefly, the problem is that, for very short
packets, Adler32 is guaranteed to give poor coverage of the available
bits. Don't take my word for it, ask Mark Adler. :-)" The problem is
that sum A does not wrap for short messages. The maximum value of A for
a 128-byte message is 32640, which is below the value 65521 used by the
modulo operation. An extended explanation can be found in RFC 3309,
which mandates the use of CRC32 instead of Adler-32 for SCTP, the
Stream Control Transmission Protocol.
In the context of a GIT pack, we have lots of small objects, especially
deltas, which are likely to be quite small and in a size range for which
Adler32 is dimed not to be sufficient. Another advantage of CRC32 is the
possibility for recovery from certain types of small corruptions like
single bit errors which are the most probable type of corruptions.
OK what this patch does is to compute the CRC32 of each object written to
a pack within pack-objects. It is not written to the index yet and it is
obviously not validated when reusing pack data yet either.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
GCC's format __attribute__ is good for checking errors, especially
with -Wformat=2 parameter. This fixes most of the reported problems
against 2005-08-09 snapshot.
Also, make the writing of the SHA1 as a end-header be conditional: not
every user will necessarily want to write the SHA1 to the file itself,
even though current users do (but we migh end up using the same helper
functions for the object files themselves, that don't do this).
This also makes the packed index file contain the SHA1 of the packed
data file at the end (just before its own SHA1). That way you can
validate the pairing of the two if you want to.
We want to be able to check their integrity later, and putting the
sha1-sum of the contents at the end is a good thing. The writing
routines are generic, so we could try to re-use them for the index file,
instead of having the same logic duplicated.
Update unpack-objects to know about the extra 20 bytes at the end
of the index.
Nicolas Pitre
Linus Torvalds
Junio C Hamano
Dana How
Timo Sirainen
Linus Torvalds