Opening the same repository multiple times will currently open the same
file multiple times, as well as map the same region of the file multiple
times. This is not necessary, as the packfile data is immutable.
Instead of opening and closing packfiles directly, introduce an
indirection and allocate packfiles globally. This does mean locking on
each packfile open, but we already use this lock for the global mwindow
list so it doesn't introduce a new contention point.
Windows has its own ftruncate() called _chsize_s().
p_mkstemp() is changed to use p_open() so we can make sure we open for
writing; the addition of exclusive create is a good thing to do
regardless, as we want a temporary path for ourselves.
Lastly, MSVC doesn't quite know how to add two numbers if one of them is a
void pointer, so let's alias it to unsigned char.C
Some OSs cannot keep their ideas about file content straight when mixing
standard IO with file mapping. As we use mmap for reading from the
packfile, let's make writing to the pack file use mmap.
Our vector does a move of the rest of the array when we remove an
item. Doing this repeatedly can be expensive, and we do this a lot in
the indexer. Instead, set the value to NULL and skip those entries.
perf reported around 30% of `index-pack` time was going into
memmove. With this change, that goes away and we spent most of the time
hashing and inflating data.
This renames git_vector_free_all to the better git_vector_free_deep
and also contains a couple of memory leak fixes based on valgrind
checks. The fixes are specifically: failure to free global dir
path variables when not compiled with threading on and failure to
free filters from the filter registry that had not be initialized
fully.
This adds tests that try canceling an indexer operation from
within the progress callback.
After writing the tests, I wanted to run this under valgrind and
had a number of errors in that situation because mmap wasn't
working. I added a CMake option to force emulation of mmap and
consolidated the Amiga-specific code into that new place (so we
don't actually need separate Amiga code now, just have to turn on
-DNO_MMAP).
Additionally, I made the indexer code propagate error codes more
reliably than it used to.
This changes the behavior of callbacks so that the callback error
code is not converted into GIT_EUSER and instead we propagate the
return value through to the caller. Instead of using the
giterr_capture and giterr_restore functions, we now rely on all
functions to pass back the return value from a callback.
To avoid having a return value with no error message, the user
can call the public giterr_set_str or some such function to set
an error message. There is a new helper 'giterr_set_callback'
that functions can invoke after making a callback which ensures
that some error message was set in case the callback did not set
one.
In places where the sign of the callback return value is
meaningful (e.g. positive to skip, negative to abort), only the
negative values are returned back to the caller, obviously, since
the other values allow for continuing the loop.
The hardest parts of this were in the checkout code where positive
return values were overloaded as meaningful values for checkout.
I fixed this by adding an output parameter to many of the internal
checkout functions and removing the overload. This added some
code, but it is probably a better implementation.
There is some funkiness in the network code where user provided
callbacks could be returning a positive or a negative value and
we want to rely on that to cancel the loop. There are still a
couple places where an user error might get turned into GIT_EUSER
there, I think, though none exercised by the tests.
There are a lot of places that we call git__free on each item in
a vector and then call git_vector_free on the vector itself. This
just wraps that up into one convenient helper function.
This continues auditing all the places where GIT_EUSER is being
returned and making sure to clear any existing error using the
new giterr_user_cancel helper. As a result, places that relied
on intercepting GIT_EUSER but having the old error preserved also
needed to be cleaned up to correctly stash and then retrieve the
actual error.
Additionally, as I encountered places where error codes were not
being propagated correctly, I tried to fix them up. A number of
those fixes are included in the this commit as well.
It was there to keep it apart from the one which read in from a file on
disk. This other indexer does not exist anymore, so there is no need for
anything other than git_indexer to refer to it.
While here, rename _add() function to _append() and _finalize() to
_commit(). The former change is cosmetic, while the latter avoids
talking about "finalizing", which OO languages use to mean something
completely different.
The user is unable to derive the number of deltas in the pack, as that
would require them to capture the stats exactly in the moment between
download and final processing, which is abstracted away in the fetch.
Capture these numbers for the user and expose them in the progress
struct. The clone and fetch examples now also present this information
to the user.
Don't increase the number of total objects, as it can produce
suprising progress output. The only addition compared to pre-thin is
the addition of local_objects to allow an output similar to git's
"completed with %d local objects".
There may be multiple deltas referencing the same base as well as OFS
deltas which rely on a thin delta. Deal with both at the same time by
injecting a single object and going back up to the main
delta-resolving loop.
When given an ODB from which to read objects, the indexer will attempt
to inject the missing bases at the end of the pack and update the
header and trailer to reflect the new contents.
Though unusual, a packfile may contain a delta whose base is a delta
that comes later. In order index such a packfile, we must not give up
on the first failure to resolve a delta, but keep it around.
If there is a pass which makes no progress, this indicates that the
packfile is broken, so fail accordingly.
The indexer code was generating warnings on Windows 64-bit. I
looked closely at the logic and was able to simplify it a bit.
Also this fixes some other Windows and Linux warnings.
This commit adds cancellation for the push operation. This work consists of:
1) Support cancellation during push operation
- During object counting phase
- During network transfer phase
- Propagate GIT_EUSER error code out to caller
2) Improve cancellation support during fetch
- Handle cancellation request during network transfer phase
- Clear error string when cancelled during indexing
3) Fix error handling in git_smart__download_pack
Cancellation during push is still only handled in the pack building and
network transfer stages of push (and not during packbuilding).
Rename git_packfile_check to git_packfile_alloc since it is now
being used more in that capacity. Fix the various places that use
it. Consolidate some repeated code in odb_pack.c related to the
allocation of a new pack_backend.
The indexer was creating a packfile object separately from the
code in pack.c which was a problem since I put a call to
git_mutex_init into just pack.c. This commit updates the pack
function for creating a new pack object (i.e. git_packfile_check())
so that it can be used in both places and then makes indexer.c
use the shared initialization routine.
There are also a few minor formatting and warning message fixes.
Implicit type conversion argument of function to size_t type
Suspicious sequence of types castings: size_t -> int -> size_t
Consider reviewing the expression of the 'A = B == C' kind. The expression is calculated as following: 'A = (B == C)'
Unsigned type is never < 0
These offsets are needed for REF_DELTA objects, which encode which
object they use as a base, but not where it lies in the packfile, so
we need a list.
These objects are mostly from older packfiles, before OFS_DELTA was
widely spread. The time spent in indexing these packfiles is greatly
reduced, though remains above what git is able to do.
This was the first implementation and its goal was simply to have
something that worked. It is slow and now it's just taking up
space. Remove it and switch the one known usage to use the streaming
indexer.
