A handful of manual pages called MERGE_AUTOSTASH a "special ref",
but there is nothing special about it. It merely is yet another
pseudoref.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A handful of manual pages called AUTO_MERGE a "special ref", but
there is nothing special about it. It merely is yet another
pseudoref.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The description of "git bisect --no-checkout" called BISECT_HEAD a
"special ref", but there is nothing special about it. It merely is
yet another pseudoref.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The introductory text in "git help git" that describes HEAD called
it "a special ref". It is special compared to the more regular refs
like refs/heads/master and refs/tags/v1.0.0, but not that special,
unlike truly special ones like FETCH_HEAD.
Rewrite a few sentences to also introduce the distinction between a
regular ref that contain the object name and a symbolic ref that
contain the name of another ref. Update the description of HEAD
that point at the current branch to use the more correct term, a
"symbolic ref".
This was found as part of auditing the documentation and in-code
comments for uses of "special ref" that refer merely a "pseudo ref".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With one exception, the synopsis for `git add` consistently lists the
short counterpart alongside the long-form of each option (for instance,
"[--edit | -e]"). The exception is that -A is not mentioned alongside
--all. Fix this inconsistency
Reported-by: Benjamin Lehmann <ben.lehmann@gmail.com>
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "check-chainlint" target runs automatically when running tests and
performs self-checks to verify that the chainlinter itself produces the
expected output. Originally, the chainlinter was implemented via sed,
but the infrastructure has been rewritten in fb41727b7e (t: retire
unused chainlint.sed, 2022-09-01) to use a Perl script instead.
The rewrite caused some slight whitespace changes in the output that are
ultimately not of much importance. In order to be able to assert that
the actual chainlinter errors match our expectations we thus have to
ignore whitespace characters when diffing them. As the `-w` flag is not
in POSIX we try to use `git diff -w --no-index` before we fall back to
`diff -w -u`.
To accomodate for cases where the host system has no Git installation we
use the locally-compiled version of Git. This can result in problems
though when the Git project's repository is using extensions that the
locally-compiled version of Git doesn't understand. It will refuse to
run and thus cause the checks to fail.
Instead of improving the detection logic, fix our ".expect" files so
that we do not need any post-processing at all anymore. This allows us
to drop the `-w` flag when diffing so that we can always use diff(1)
now.
Note that we keep some of the post-processing of `chainlint.pl` output
intact to strip leading line numbers generated by the script. Having
these would cause a rippling effect whenever we add a new test that
sorts into the middle of existing tests and would require us to
renumerate all subsequent lines, which seems rather pointless.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Reviewed-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To ensure that we don't regress either the size or runtime performance
of multi-pack reuse, add a performance test to measure both of these.
The test partitions the objects in GIT_TEST_PERF_LARGE_REPO into 1, 10,
and 100 packs, and then tries to perform a "clone" at each stage with
both single- and multi-pack reuse enabled.
Note that the `repack_into_n_chunks()` function in this new test script
differs from the existing `repack_into_n()`. The former partitions the
repository into N equal-sized chunks, while the latter produces N packs
of five commits each (plus their objects), and then another pack with
the remainder.
On git.git, I can produce the following results on my machine:
Test this tree
--------------------------------------------------------------------------------
5332.3: clone for 1-pack scenario (single-pack reuse) 1.57(2.99+0.15)
5332.4: clone size for 1-pack scenario (single-pack reuse) 231.8M
5332.5: clone for 1-pack scenario (multi-pack reuse) 1.79(2.96+0.21)
5332.6: clone size for 1-pack scenario (multi-pack reuse) 231.7M
5332.9: clone for 10-pack scenario (single-pack reuse) 3.89(16.75+0.35)
5332.10: clone size for 10-pack scenario (single-pack reuse) 209.9M
5332.11: clone for 10-pack scenario (multi-pack reuse) 1.56(2.99+0.17)
5332.12: clone size for 10-pack scenario (multi-pack reuse) 224.4M
5332.15: clone for 100-pack scenario (single-pack reuse) 8.24(54.31+0.59)
5332.16: clone size for 100-pack scenario (single-pack reuse) 278.3M
5332.17: clone for 100-pack scenario (multi-pack reuse) 2.13(2.44+0.33)
5332.18: clone size for 100-pack scenario (multi-pack reuse) 357.9M
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that both the pack-bitmap and pack-objects code are prepared to
handle marking and using objects from multiple bitmapped packs for
verbatim reuse, allow marking objects from all bitmapped packs as
eligible for reuse.
Within the `reuse_partial_packfile_from_bitmap()` function, we no longer
only mark the pack whose first object is at bit position zero for reuse,
and instead mark any pack contained in the MIDX as a reuse candidate.
Provide a handful of test cases in a new script (t5332) exercising
interesting behavior for multi-pack reuse to ensure that we performed
all of the previous steps correctly.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In e704fc7978 (pack-objects: introduce pack.allowPackReuse, 2019-12-18),
the `pack.allowPackReuse` configuration option was introduced, allowing
users to disable the pack reuse mechanism.
To prepare for debugging multi-pack reuse, allow setting configuration
to "single" in addition to the usual bool-or-int values.
"single" implies the same behavior as "true", "1", "yes", and so on. But
it will complement a new "multi" value (to be introduced in a future
commit). When set to "single", we will only perform pack reuse on a
single pack, regardless of whether or not there are multiple MIDX'd
packs.
This requires no code changes (yet), since we only support single pack
reuse.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce a helper function which looks for a specific (category, key,
value) tuple in the output of a trace2 event stream.
We will use this function in a future patch to ensure that the expected
number of objects are reused from an expected number of packs.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As part of the multi-pack reuse effort, we will want to add some tests
that assert that we reused a certain number of objects from a certain
number of packs.
We could do this by grepping through the stderr output of
`pack-objects`, but doing so would be brittle in case the output format
changed.
Instead, let's use the trace2 mechanism to log various pieces of
information about the generated packfile, which we can then use to
compare against desired values.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the pack-objects code is equipped to handle reusing objects
from multiple packs, prepare the pack-bitmap code to mark objects from
multiple packs as reuse candidates.
In order to prepare the pack-bitmap code for this change, remove the
same set of assumptions we unwound in previous commits from the helper
function `reuse_partial_packfile_from_bitmap_1()`, in preparation for it
to be called in a loop over the set of bitmapped packs in a following
commit.
Most importantly, we can no longer assume that the bit position
corresponding to the first object in a given reuse pack candidate is at
the beginning of the bitmap itself.
For the single pack that this assumption is still true for (in MIDX
bitmaps, this is the preferred pack, in single-pack bitmaps it is the
pack the bitmap is tied to), we can still use our whole-words
optimization.
But for all subsequent packs, we can not make use of this optimization,
since it assumes that all delta bases are being sent from the same pack,
which would break if we are sending OFS_DELTAs down to the client. To
understand why, consider two packs, P1 and P2 where:
- P1 has object A which is a delta on base B
- P2 has its own copy of B, in addition to other objects
Suppose that the MIDX which covers P1 and P2 selected its copy of A from
P1, but selected its copy of B from P2. Since A is a delta of B, but the
base was selected from a different pack, sending the bytes corresponding
to A as an OFS_DELTA verbatim from P1 would be incorrect, since we don't
guarantee that B is in the same place relative to A in the generated
pack as in P1.
For now, we detect and reject these cross-pack deltas by searching for
the (pack_id, offset) pair for the delta's base object (using the same
pack_id as the pack containing the delta'd object) in the MIDX. If we
find a match, that means that the MIDX did indeed pick the base object
from the same pack, and we are OK to reuse the delta.
If we don't find a match, however, that means that the base object was
selected from a different pack in the MIDX, and we can let the slower
path handle re-delta'ing our candidate object.
In the future, there are a couple of other things we could do, namely:
- Turn any cross-pack deltas (which are stored as OFS_DELTAs) into
REF_DELTAs. We already do this today when reusing an OFS_DELTA
without `--delta-base-offset` enabled, so it's not a huge stretch to
do the same for cross-pack deltas even when `--delta-base-offset` is
enabled.
This would work, but would obviously result in larger-than-necessary
packs, as we in theory *could* represent these cross-pack deltas by
patching an existing OFS_DELTA. But it's not clear how much that
would matter in practice. I suspect it would have a lot to do with
how you pack your repository in the first place.
- Finally, we could patch OFS_DELTAs across packs in a similar fashion
as we do today for OFS_DELTAs within a single pack on either side of
a gap. This would result in the smallest packs of the three options
here, but implementing this would be more involved.
At minimum, you'd have to keep the reusable chunks list for all
reused packs, not just the one we're currently processing. And you'd
have to ensure that any bases which are a part of cross-pack deltas
appear before the delta. I think this is possible to do, but would
require assembling the reusable chunks list potentially in a
different order than they appear in the source packs.
For now, let's pursue the simplest approach and reject any cross-pack
deltas.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we have extracted the `midx_key_to_pack_pos()` function, we can
implement the `midx_pair_to_pack_pos()` function which accepts (pack_id,
offset) tuples and returns an index into the psuedo-pack order.
This will be used in a following commit in order to figure out whether
or not the MIDX chose a given delta's base object from the same pack as
the delta resides in. It will do so by locating the base object's offset
in the pack, and then performing a binary search using the same pack ID
with the base object's offset.
If (and only if) it finds a match (at any position) we can guarantee
that the MIDX selected both halves of the delta/base pair from the same
pack.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `midx_to_pack_pos()` function implements a binary search over
objects in the MIDX between lexical and pseudo-pack order. It does this
by taking in an index into the lexical order (i.e. the same argument
you'd use for `nth_midxed_object_id()` and similar) and spits out a
position in the pseudo-pack order.
This works for all callers, since they currently all are translating
from lexical order to pseudo-pack order. But future callers may want to
translate a known (offset, pack_id) tuple into an index into the
psuedo-pack order, without knowing where that (offset, pack_id) tuple
appears in lexical order.
Prepare for implementing a function that translates between a (offset,
pack_id) tuple into an index into the psuedo-pack order by extracting a
helper function which does just that, and then reimplementing
midx_to_pack_pos() in terms of it.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When performing a binary search over the objects in a MIDX's bitmap
(i.e. in pseudo-pack order), the reader reconstructs the pseudo-pack
ordering using a combination of (a) the preferred pack, (b) the pack's
lexical position in the MIDX based on pack names, and (c) the object
offset within the pack.
In order to perform this binary search, the reader must know the
identity of the preferred pack. This could be stored in the MIDX, but
isn't for historical reasons, mostly because it can easily be inferred
at read-time by looking at the object in the first bit position and
finding out which pack it was selected from in the MIDX, like so:
nth_midxed_pack_int_id(m, pack_pos_to_midx(m, 0));
In midx_to_pack_pos() which performs this binary search, we look up the
identity of the preferred pack before each search. This is relatively
quick, since it involves two table-driven lookups (one in the MIDX's
revindex for `pack_pos_to_midx()`, and another in the MIDX's object
table for `nth_midxed_pack_int_id()`).
But since the preferred pack does not change after the MIDX is written,
it is safe to cache this value on the MIDX itself.
Write a helper to do just that, and rewrite all of the existing
call-sites that care about the identity of the preferred pack in terms
of this new helper.
This will prepare us for a subsequent patch where we will need to binary
search through the MIDX's pseudo-pack order multiple times.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar fashion as other checked cast functions in this header
(such as `cast_size_t_to_ulong()` and `cast_size_t_to_int()`), implement
a checked cast function for going from a size_t to a uint32_t value.
This function will be utilized in a future commit which needs to make
such a conversion.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In addition to including the number of objects reused verbatim from a
reuse-pack, include the number of packs from which objects were reused.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `write_reused_pack_verbatim()` within
`builtin/pack-objects.c` is responsible for writing out a continuous
set of objects beginning at the start of the reuse packfile.
In the existing implementation, we did something like:
while (pos < reuse_packfile_bitmap->word_alloc &&
reuse_packfile_bitmap->words[pos] == (eword_t)~0)
pos++;
if (pos)
/* write first `pos * BITS_IN_WORD` objects from pack */
as an optimization to record a single chunk for the longest continuous
prefix of objects wanted out of the reuse pack, instead of having a
chunk for each individual object. For more details, see bb514de356
(pack-objects: improve partial packfile reuse, 2019-12-18).
In order to retain this optimization in a multi-pack reuse world, we can
no longer assume that the first object in a pack is on a word boundary
in the bitmap storing the set of reusable objects.
Assuming that all objects from the beginning of the reuse packfile up to
the object corresponding to the first bit on a word boundary are part of
the result, consume whole words at a time until the last whole word
belonging to the reuse packfile. Copy those objects to the resulting
packfile, and track that we reused them by recording a single chunk.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `write_reused_pack()` within `builtin/pack-objects.c` is
responsible for performing pack-reuse on a single pack, and has two main
functions:
- it dispatches a call to `write_reused_pack_verbatim()` to see if we
can reuse portions of the packfile in whole-word chunks
- for any remaining objects (that is, any objects that appear after
the first "gap" in the bitmap), call write_reused_pack_one() on that
object to record it for reuse.
Prepare this function for multi-pack reuse by removing the assumption
that the bit position corresponding to the first object being reused
from a given pack must be at bit position zero.
The changes in this function are mostly straightforward. Initialize `i`
to the position of the first word to contain bits corresponding to that
reuse pack. In most situations, we throw the initialized value away,
since we end up replacing it with the return value from
write_reused_pack_verbatim(), moving us past the section of whole words
that we reused.
Likewise, modify the per-object loop to ignore any bits at the beginning
of the first word that do not belong to the pack currently being reused,
as well as skip to the "done" section once we have processed the last
bit corresponding to this pack.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Further prepare pack-objects to perform verbatim pack-reuse over
multiple packfiles by converting functions that take in a pointer to a
`struct packed_git` to instead take in a pointer to a `struct
bitmapped_pack`.
The additional information found in the bitmapped_pack struct (such as
the bit position corresponding to the beginning of the pack) will be
necessary in order to perform verbatim pack-reuse.
Note that we don't use any of the extra pieces of information contained
in the bitmapped_pack struct, so this step is merely preparatory and
does not introduce any functional changes.
Note further that we do not change the argument type to
write_reused_pack_one(). That function is responsible for copying
sections of the packfile directly and optionally patching any OFS_DELTAs
to account for not reusing sections of the packfile in between a delta
and its base.
As such, that function is (and should remain) oblivious to multi-pack
reuse, and does not require any of the extra pieces of information
stored in the bitmapped_pack struct.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When reusing objects from a pack, we keep track of a set of one or more
`reused_chunk`s, corresponding to sections of one or more object(s) from
a source pack that we are reusing. Each chunk contains two pieces of
information:
- the offset of the first object in the source pack (relative to the
beginning of the source pack)
- the difference between that offset, and the corresponding offset in
the pack we're generating
The purpose of keeping track of these is so that we can patch an
OFS_DELTAs that cross over a section of the reuse pack that we didn't
take.
For instance, consider a hypothetical pack as shown below:
(chunk #2)
__________...
/
/
+--------+---------+-------------------+---------+
... | <base> | <other> | (unused) | <delta> | ...
+--------+---------+-------------------+---------+
\ /
\______________/
(chunk #1)
Suppose that we are sending objects "base", "other", and "delta", and
that the "delta" object is stored as an OFS_DELTA, and that its base is
"base". If we don't send any objects in the "(unused)" range, we can't
copy the delta'd object directly, since its delta offset includes a
range of the pack that we didn't copy, so we have to account for that
difference when patching and reassembling the delta.
In order to compute this value correctly, we need to know not only where
we are in the packfile we're assembling (with `hashfile_total(f)`) but
also the position of the first byte of the packfile that we are
currently reusing. Currently, this works just fine, since when reusing
only a single pack those two values are always identical (because
verbatim reuse is the first thing pack-objects does when enabled after
writing the pack header).
But when reusing multiple packs which have one or more gaps, we'll need
to account for these two values diverging.
Together, these two allow us to compute the reused chunk's offset
difference relative to the start of the reused pack, as desired.
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The routines pack-objects uses to perform verbatim pack-reuse are:
- write_reused_pack_one()
- write_reused_pack_verbatim()
- write_reused_pack()
, all of which assume that there is exactly one packfile being reused:
the global constant `reuse_packfile`.
Prepare for reusing objects from multiple packs by making reuse packfile
a parameter of each of the above functions in preparation for calling
these functions in a loop with multiple packfiles.
Note that we still have the global "reuse_packfile", but pass it through
each of the above function's parameter lists, eliminating all but one
direct access (the top-level caller in `write_pack_file()`). Even after
this series, we will still have a global, but it will hold the array of
reusable packfiles, and we'll pass them one at a time to these functions
in a loop.
Note also that we will eventually need to pass a `bitmapped_pack`
instead of a `packed_git` in order to hold onto additional information
required for reuse (such as the bit position of the first object
belonging to that pack). But that change will be made in a future commit
so as to minimize the noise below as much as possible.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Further prepare for enabling verbatim pack-reuse over multiple packfiles
by changing the signature of reuse_partial_packfile_from_bitmap() to
populate an array of `struct bitmapped_pack *`'s instead of a pointer to
a single packfile.
Since the array we're filling out is sized dynamically[^1], add an
additional `size_t *` parameter which will hold the number of reusable
packs (equal to the number of elements in the array).
Note that since we still have not implemented true multi-pack reuse,
these changes aren't propagated out to the rest of the caller in
builtin/pack-objects.c.
In the interim state, we expect that the array has a single element, and
we use that element to fill out the static `reuse_packfile` variable
(which is a bog-standard `struct packed_git *`). Future commits will
continue to push this change further out through the pack-objects code.
[^1]: That is, even though we know the number of packs which are
candidates for pack-reuse, we do not know how many of those
candidates we can actually reuse.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The signature of `reuse_partial_packfile_from_bitmap()` currently takes
in a bitmap, as well as three output parameters (filled through
pointers, and passed as arguments), and also returns an integer result.
The output parameters are filled out with: (a) the packfile used for
pack-reuse, (b) the number of objects from that pack that we can reuse,
and (c) a bitmap indicating which objects we can reuse. The return value
is either -1 (when there are no objects to reuse), or 0 (when there is
at least one object to reuse).
Some of these parameters are redundant. Notably, we can infer from the
bitmap how many objects are reused by calling bitmap_popcount(). And we
can similar compute the return value based on that number as well.
As such, clean up the signature of this function to drop the "*entries"
parameter, as well as the int return value, since the single caller of
this function can infer these values themself.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a future commit, we will want to check whether or not a bitmap has
any bits set in any of its words. The best way to do this (prior to the
existence of this patch) is to call `bitmap_popcount()` and check
whether the result is non-zero.
But this is semi-wasteful, since we do not need to know the exact number
of bits set, only whether or not there is at least one of them.
Implement a new helper function to check just that.
Suggested-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When trying to assemble a pack with bitmaps using `--use-bitmap-index`,
`pack-objects` asks the pack-bitmap machinery for a bitmap which
indicates the set of objects we can "reuse" verbatim from on-disk.
This set is roughly comprised of: a prefix of objects in the bitmapped
pack (or preferred pack, in the case of a multi-pack reachability
bitmap), plus any other objects not included in the prefix, excluding
any deltas whose base we are not sending in the resulting pack.
The pack-bitmap machinery is responsible for computing this bitmap, and
does so with the following functions:
- reuse_partial_packfile_from_bitmap()
- try_partial_reuse()
In the existing implementation, the first function is responsible for
(a) marking the prefix of objects in the reusable pack, and then (b)
calling try_partial_reuse() on any remaining objects to ensure that they
are also reusable (and removing them from the bitmapped set if they are
not).
Likewise, the `try_partial_reuse()` function is responsible for checking
whether an isolated object (that is, an object from the bitmapped
pack/preferred pack not contained in the prefix from earlier) may be
reused, i.e. that it isn't a delta of an object that we are not sending
in the resulting pack.
These functions are based on two core assumptions, which we will unwind
in this and the following commits:
1. There is only a single pack from the bitmap which is eligible for
verbatim pack-reuse. For single-pack bitmaps, this is trivially the
bitmapped pack. For multi-pack bitmaps, this is (currently) the
MIDX's preferred pack.
2. The pack eligible for reuse has its first object in bit position 0,
and all objects from that pack follow in pack-order from that first
bit position.
In order to perform verbatim pack reuse over multiple packs, we must
unwind these two assumptions. Most notably, in order to reuse bits from
a given packfile, we need to know the first bit position occupied by
an object form that packfile. To propagate this information around, pass
a `struct bitmapped_pack *` anywhere we previously passed a `struct
packed_git *`, since the former contains the bitmap position we're
interested in (as well as a pointer to the latter).
As an additional step, factor out a sub-routine from the main
`reuse_partial_packfile_from_bitmap()` function, called
`reuse_partial_packfile_from_bitmap_1()`. This new function will be
responsible for figuring out which objects may be reused from a single
pack, and the existing function will dispatch multiple calls to its new
helper function for each reusable pack.
Consequently, `reuse_partial_packfile_from_bitmap()` will now maintain
an array of reusable packs instead of a single such pack. We currently
expect that array to have only a single element, so this awkward state
is short-lived. It will serve as useful scaffolding in subsequent
commits as we begin to work towards enabling multi-pack reuse.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The multi-pack index API exposes a `midx_contains_pack()` function that
takes in a string ending in either ".idx" or ".pack" and returns whether
or not the MIDX contains a given pack corresponding to that string.
There is no corresponding function to locate the position of a pack
within the MIDX's pack order (sorted lexically by pack filename).
We could add an optional out parameter to `midx_contains_pack()` that is
filled out with the pack's position when the parameter is non-NULL. To
minimize the amount of fallout from this change, instead introduce a new
function by renaming `midx_contains_pack()` to `midx_locate_pack()`,
adding that output parameter, and then reimplementing
`midx_contains_pack()` in terms of it.
Future patches will make use of this new function.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a multi-pack bitmap is used to implement verbatim pack reuse (that
is, when verbatim chunks from an on-disk packfile are copied
directly[^1]), it does so by using its "preferred pack" as the source
for pack-reuse.
This allows repositories to pack the majority of their objects into a
single (often large) pack, and then use it as the single source for
verbatim pack reuse. This increases the amount of objects that are
reused verbatim (and consequently, decrease the amount of time it takes
to generate many packs). But this performance comes at a cost, which is
that the preferred packfile must pace its growth with that of the entire
repository in order to maintain the utility of verbatim pack reuse.
As repositories grow beyond what we can reasonably store in a single
packfile, the utility of verbatim pack reuse diminishes. Or, at the very
least, it becomes increasingly more expensive to maintain as the pack
grows larger and larger.
It would be beneficial to be able to perform this same optimization over
multiple packs, provided some modest constraints (most importantly, that
the set of packs eligible for verbatim reuse are disjoint with respect
to the subset of their objects being sent).
If we assume that the packs which we treat as candidates for verbatim
reuse are disjoint with respect to any of their objects we may output,
we need to make only modest modifications to the verbatim pack-reuse
code itself. Most notably, we need to remove the assumption that the
bits in the reachability bitmap corresponding to objects from the single
reuse pack begin at the first bit position.
Future patches will unwind these assumptions and reimplement their
existing functionality as special cases of the more general assumptions
(e.g. that reuse bits can start anywhere within the bitset, but happen
to start at 0 for all existing cases).
This patch does not yet relax any of those assumptions. Instead, it
implements a foundational data-structure, the "Bitampped Packs" (`BTMP`)
chunk of the multi-pack index. The `BTMP` chunk's contents are described
in detail here. Importantly, the `BTMP` chunk contains information to
map regions of a multi-pack index's reachability bitmap to the packs
whose objects they represent.
For now, this chunk is only written, not read (outside of the test-tool
used in this patch to test the new chunk's behavior). Future patches
will begin to make use of this new chunk.
[^1]: Modulo patching any `OFS_DELTA`'s that cross over a region of the
pack that wasn't used verbatim.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When selecting which packfiles will be written while generating a MIDX,
the MIDX internals fill out a 'struct pack_info' with various pieces of
book-keeping.
Instead of filling out each field of the `pack_info` structure
individually in each of the two spots that modify the array of such
structures (`ctx->info`), extract a common routine that does this for
us.
This reduces the code duplication by a modest amount. But more
importantly, it zero-initializes the structure before assigning values
into it. This hardens us for a future change which will add additional
fields to this structure which (until this patch) was not
zero-initialized.
As a result, any new fields added to the `pack_info` structure need only
be updated in a single location, instead of at each spot within midx.c.
There are no functional changes in this patch.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `find_objects()` function creates an object_list for any tips of the
reachability query which do not have corresponding bitmaps.
The object_list is not used outside of `find_objects()`, but we never
free it with `object_list_free()`, resulting in a leak. Let's plug that
leak by calling `object_list_free()`, which results in t6113 becoming
leak-free.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `bb_commit` commit slab is used by the pack-bitmap-write machinery
to track various pieces of bookkeeping used to generate reachability
bitmaps.
Even though we clear the slab when freeing the bitmap_builder struct
(with `bitmap_builder_clear()`), there are still pointers which point to
locations in memory that have not yet been freed, resulting in a leak.
Plug the leak by introducing a suitable `free_fn` for the `struct
bb_commit` type, and make sure it is called on each member of the slab
via the `deep_clear_bb_data()` function.
Note that it is possible for both of the arguments to `bitmap_free()` to
be NULL, but `bitmap_free()` is a noop for NULL arguments, so it is OK
to pass them unconditionally.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The pack-objects internals use a packing_data struct to track what
objects are part of the pack(s) being formed.
Since these structures contain allocated fields, failing to
appropriately free() them results in a leak. Plug that leak by
introducing a clear_packing_data() function, and call it in the
appropriate spots.
This is a fairly straightforward leak to plug, since none of the callers
expect to read any values or have any references to parts of the address
space being freed.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our unquote_comment() function is recursive; when it sees a comment
within a comment, like:
(this is an (embedded) comment)
it recurses to handle the inner comment. This is fine for practical use,
but it does mean that you can easily run out of stack space with a
malicious header. For example:
perl -e 'print "From: ", "(" x 2**18;' |
git mailinfo /dev/null /dev/null
segfaults on my system. And since mailinfo is likely to be fed untrusted
input from the Internet (if not by human users, who might recognize a
garbage header, but certainly there are automated systems that apply
patches from a list) it may be possible for an attacker to trigger the
problem.
That said, I don't think there's an interesting security vulnerability
here. All an attacker can do is make it impossible to parse their email
and apply their patch, and there are lots of ways to generate bogus
emails. So it's more of an annoyance than anything.
But it's pretty easy to fix it. The recursion is not helping us preserve
any particular state from each level. The only flag in our parsing is
take_next_literally, and we can never recurse when it is set (since the
start of a new comment implies it was not backslash-escaped). So it is
really only useful for finding the end of the matched pair of
parentheses. We can do that easily with a simple depth counter.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When processing "From" headers in an email, mailinfo "unquotes" quoted
strings and rfc822 parenthesized comments. For quoted strings, we
actually remove the double-quotes, so:
From: "A U Thor" <someone@example.com>
become:
Author: A U Thor
Email: someone@example.com
But for comments, we leave the outer parentheses in place, so:
From: A U (this is a comment) Thor <someone@example.com>
becomes:
Author: A U (this is a comment) Thor
Email: someone@example.com
So what is the comment "unquoting" actually doing? In our code, being in
a comment section has exactly two effects:
1. We'll unquote backslash-escaped characters inside a comment
section.
2. We _won't_ unquote double-quoted strings inside a comment section.
Our test for comments in t5100 checks this:
From: "A U Thor" <somebody@example.com> (this is \(really\) a comment (honestly))
So it is covering (1), but not (2). Let's add in a quoted string to
cover this.
Moreover, because the comment appears at the end of the From header,
there's nothing to confirm that we correctly found the end of the
comment section (and not just the end-of-string). Let's instead move it
to the beginning of the header, which means we can confirm that the
existing quoted string is detected (which will only happen if we know
we've left the comment block).
As expected, the test continues to pass, but this will give us more
confidence as we refactor the code in the next patch.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're inconsistently writing BISECT_EXPECTED_REV both via the filesystem
and via the refdb, which violates the newly established rules for how
special refs must be treated. This works alright in practice with the
reffiles reference backend, but will cause bugs once we gain additional
backends.
Fix this issue and consistently write BISECT_EXPECTED_REV via the refdb
so that it is no longer a special ref.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We have some references that are more special than others. The reason
for them being special is that they either do not follow the usual
format of references, or that they are written to the filesystem
directly by the respective owning subsystem and thus circumvent the
reference backend.
This works perfectly fine right now because the reffiles backend will
know how to read those refs just fine. But with the prospect of gaining
a new reference backend implementation we need to be a lot more careful
here:
- We need to make sure that we are consistent about how those refs are
written. They must either always be written via the filesystem, or
they must always be written via the reference backend. Any mixture
will lead to inconsistent state.
- We need to make sure that such special refs are always handled
specially when reading them.
We're already mostly good with regard to the first item, except for
`BISECT_EXPECTED_REV` which will be addressed in a subsequent commit.
But the current list of special refs is missing some refs that really
should be treated specially. Right now, we only treat `FETCH_HEAD` and
`MERGE_HEAD` specially here.
Introduce a new function `is_special_ref()` that contains all current
instances of special refs to fix the reading path.
Note that this is only a temporary measure where we record and rectify
the current state. Ideally, the list of special refs should in the end
only contain `FETCH_HEAD` and `MERGE_HEAD` again because they both may
reference multiple objects and can contain annotations, so they indeed
are special.
Based-on-patch-by: Han-Wen Nienhuys <hanwenn@gmail.com>
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some refs in Git are more special than others due to reasons explained
in the next commit. These refs are read via `refs_read_special_head()`,
but this function doesn't behave the same as when we try to read a
normal ref. Most importantly, we do not propagate `failure_errno` in the
case where the reference does not exist, which is behaviour that we rely
on in many parts of Git.
Fix this bug by propagating errno when `strbuf_read_file()` fails.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We read both the HEAD and ORIG_HEAD references directly from the
filesystem in order to figure out whether we're currently splitting a
commit. If both of the following are true:
- HEAD points to the same object as "rebase-merge/amend".
- ORIG_HEAD points to the same object as "rebase-merge/orig-head".
Then we are currently splitting commits.
The current code only works by chance because we only have a single
reference backend implementation. Refactor it to instead read both refs
via the refdb layer so that we'll also be compatible with alternate
reference backends.
There are some subtleties involved here:
- We pass `RESOLVE_REF_READING` so that a missing ref will cause
`read_ref_full()` to return an error.
- We pass `RESOLVE_REF_NO_RECURSE` so that we do not try to resolve
symrefs. The old code didn't resolve symrefs either, and we only
ever write object IDs into the refs in "rebase-merge/".
- In the same spirit we verify that successfully-read refs are not
symbolic refs.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git checkout -B <branch> [<start-point>]", being a "forced" version
of "-b", switches to the <branch>, after optionally resetting its
tip to the <start-point>, even if the <branch> is in use in another
worktree, which is somewhat unexpected.
Protect the <branch> using the same logic that forbids "git checkout
<branch>" from touching a branch that is in use elsewhere.
This is a breaking change that may deserve backward compatibliity
warning in the Release Notes. The "--ignore-other-worktrees" option
can be used as an escape hatch if the finger memory of existing
users depend on the current behaviour of "-B".
Reported-by: Willem Verstraeten <willem.verstraeten@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
f4ee22b526 (ref-filter: add tests for objectsize:disk, 2018-12-24)
hard-coded the expected object sizes. Coincidentally the size of commit
and tag is the same with zlib at the default compression level.
1f5f8f3e85 (t6300: abstract away SHA-1-specific constants, 2020-02-22)
encoded the sizes as a single value, which coincidentally also works
with sha256.
Different compression libraries like zlib-ng may arrive at different
values. Get them from the file system instead of hard-coding them to
make switching the compression library (or changing the compression
level) easier.
Reported-by: Ondrej Pohorelsky <opohorel@redhat.com>
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When processing a header like a "From" line, mailinfo uses
unquote_quoted_pair() to handle double-quotes and rfc822 parenthesized
comments. It takes a NUL-terminated string on input, and loops over the
"in" pointer until it sees the NUL. When it finds the start of an
interesting block, it delegates to helper functions which also increment
"in", and return the updated pointer.
But there's a bug here: the helpers find the NUL with a post-increment
in the loop condition, like:
while ((c = *in++) != 0)
So when they do see a NUL (rather than the correct termination of the
quote or comment section), they return "in" as one _past_ the NUL
terminator. And thus the outer loop in unquote_quoted_pair() does not
realize we hit the NUL, and keeps reading past the end of the buffer.
We should instead make sure to return "in" positioned at the NUL, so
that the caller knows to stop their loop, too. A hacky way to do this is
to return "in - 1" after leaving the inner loop. But a slightly cleaner
solution is to avoid incrementing "in" until we are sure it contained a
non-NUL byte (i.e., doing it inside the loop body).
The two tests here show off the problem. Since we check the output,
they'll _usually_ report a failure in a normal build, but it depends on
what garbage bytes are found after the heap buffer. Building with
SANITIZE=address reliably notices the problem. The outcome (both the
exit code and the exact bytes) are just what Git happens to produce for
these cases today, and shouldn't be taken as an endorsement. It might be
reasonable to abort on an unterminated string, for example. The priority
for this patch is fixing the out-of-bounds memory access.
Reported-by: Carlos Andrés Ramírez Cataño <antaigroupltda@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're currently creating the reference database with a potentially
incorrect object format when the remote repository's object format is
different from the local default object format. This works just fine for
now because the files backend never records the object format anywhere.
But this logic will fail with any new reference backend that encodes
this information in some form either on-disk or in-memory.
The preceding commits have reshuffled code in git-clone(1) so that there
is no code path that will access the reference database before we have
detected the remote's object format. With these refactorings we can now
defer initialization of the reference database until after we have
learned the remote's object format and thus initialize it with the
correct format from the get-go.
These refactorings are required to make git-clone(1) work with the
upcoming reftable backend when cloning repositories with the SHA256
object format.
This change breaks a test in "t5550-http-fetch-dumb.sh" when cloning an
empty repository with `GIT_TEST_DEFAULT_HASH=sha256`. The test expects
the resulting hash format of the empty cloned repository to match the
default hash, but now we always end up with a sha1 repository. The
problem is that for dumb HTTP fetches, we have no easy way to figure out
the remote's hash function except for deriving it based on the hash
length of refs in `info/refs`. But as the remote repository is empty we
cannot rely on this detection mechanism.
Before the change in this commit we already initialized the repository
with the default hash function and then left it as-is. With this patch
we always use the hash function detected via the remote, where we fall
back to "sha1" in case we cannot detect it.
Neither the old nor the new behaviour are correct as we second-guess the
remote hash function in both cases. But given that this is a rather
unlikely edge case (we use the dumb HTTP protocol, the remote repository
uses SHA256 and the remote repository is empty), let's simply adapt the
test to assert the new behaviour. If we want to properly address this
edge case in the future we will have to extend the dumb HTTP protocol so
that we can properly detect the hash function for empty repositories.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After we have set up the remote configuration in git-clone(1) we'll call
`remote_get()` to read the remote from the on-disk configuration. But
next to reading the on-disk configuration, `remote_get()` will also
cause us to try and read the repository's HEAD reference so that we can
figure out the current branch. Besides being pointless in git-clone(1)
because we're operating in an empty repository anyway, this will also
break once we move creation of the reference database to a later point
in time.
Refactor the code to introduce a new `remote_get_early()` function that
will skip reading the HEAD reference to address this issue.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When asked to do a sparse checkout, then git-clone(1) will spawn
`git sparse-checkout set` to set up the configuration accordingly. This
requires a proper Git repository or otherwise the command will fail. But
as we are about to move creation of the reference database to a later
point, this prerequisite will not hold anymore.
Move the logic to a later point in time where we know to have created
the reference database already.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We create the reference database in git-clone(1) quite early before
connecting to the remote repository. Given that we do not yet know about
the object format that the remote repository uses at that point in time
the consequence is that the refdb may be initialized with the wrong
object format.
This is not a problem in the context of the files backend as we do not
encode the object format anywhere, and furthermore the only reference
that we write between initializing the refdb and learning about the
object format is the "HEAD" symref. It will become a problem though once
we land the reftable backend, which indeed does require to know about
the proper object format at the time of creation. We thus need to
rearrange the logic in git-clone(1) so that we only initialize the refdb
once we have learned about the actual object format.
As a first step, move listing of remote references to happen earlier,
which also allow us to set up the hash algorithm of the repository
earlier now. While we aim to execute this logic as late as possible
until after most of the setup has happened already, detection of the
object format and thus later the setup of the reference database must
happen before any other logic that may spawn Git commands or otherwise
these Git commands may not recognize the repository as such.
The first Git step where we expect the repository to be fully initalized
is when we fetch bundles via bundle URIs. Funny enough, the comments
there also state that "the_repository must match the cloned repo", which
is indeed not necessarily the case for the hash algorithm right now. So
in practice it is the right thing to detect the remote's object format
before downloading bundle URIs anyway, and not doing so causes clones
with bundle URIs to fail when the local default object format does not
match the remote repository's format.
Unfortunately though, this creates a new issue: downloading bundles may
take a long time, so if we list refs beforehand they might've grown
stale meanwhile. It is not clear how to solve this issue except for a
second reference listing though after we have downloaded the bundles,
which may be an expensive thing to do.
Arguably though, it's preferable to have a staleness issue compared to
being unable to clone a repository altogether.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The reftable format encodes the hash function used by the repository
inside of its tables. The reftable backend thus needs to be initialized
with the correct hash function right from the start, or otherwise we may
end up writing tables with the wrong hash function. But git-clone(1)
initializes the reference database before learning about the hash
function used by the remote repository, which has never been a problem
with the reffiles backend.
To fix this, we'll have to change git-clone(1) to be more careful and
only create the reference backend once it learned about the remote hash
function. This creates a problem for git-remote-curl(1), which will then
be spawned at a time where the repository is not yet fully-initialized.
Consequentially, git-remote-curl(1) will fail to detect the repository,
which eventually causes it to error out once it is asked to fetch remote
objects.
We can address this issue by trying to re-discover the Git repository in
case none was detected at startup time. With this change, the clone will
look as following:
1. git-clone(1) sets up the initial repository, excluding the
reference database.
2. git-clone(1) spawns git-remote-curl(1), which will be unable to
detect the repository due to a missing "HEAD".
3. git-clone(1) asks git-remote-curl(1) to list remote references.
This works just fine as this step does not require a local
repository
4. git-clone(1) creates the reference database as it has now learned
about the hash function.
5. git-clone(1) asks git-remote-curl(1) to fetch the remote packfile.
The latter notices that it doesn't have a repository available, but
it now knows to try and re-discover it.
If the re-discovery succeeds in the last step we can continue with the
clone.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allow callers to skip creation of the reference database via a new flag
`INIT_DB_SKIP_REFDB`, which is required for git-clone(1) so that we can
create it at a later point once the object format has been discovered
from the remote repository.
Note that we also uplift the call to `create_reference_database()` into
`init_db()`, which makes it easier to handle the new flag for us. This
changes the order in which we do initialization so that we now set up
the Git configuration before we create the reference database. In
practice this move should not result in any change in behaviour.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We're about to let callers skip creation of the reference database when
calling `init_db()`. Extract the logic into a standalone function so
that it becomes easier to do this refactoring.
While at it, expand the comment that explains why we always create the
"refs/" directory.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When iterating over entries in the block iterator we compute the key of
each of the entries and write it into a buffer. We do not reuse the
buffer though and thus re-allocate it on every iteration, which is
wasteful.
Refactor the code to reuse the buffer.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano