"git pack-objects" learned another algorithm to compute the set of
objects to send, that trades the resulting packfile off to save
traversal cost to favor small pushes.
* ds/push-sparse-tree-walk:
pack-objects: create GIT_TEST_PACK_SPARSE
pack-objects: create pack.useSparse setting
revision: implement sparse algorithm
list-objects: consume sparse tree walk
revision: add mark_tree_uninteresting_sparse
The code to walk tree objects has been taught that we may be
working with object names that are not computed with SHA-1.
* bc/tree-walk-oid:
cache: make oidcpy always copy GIT_MAX_RAWSZ bytes
tree-walk: store object_id in a separate member
match-trees: use hashcpy to splice trees
match-trees: compute buffer offset correctly when splicing
tree-walk: copy object ID before use
When creating a pack-file using 'git pack-objects --revs' we provide
a list of interesting and uninteresting commits. For example, a push
operation would make the local topic branch be interesting and the
known remote refs as uninteresting. We want to discover the set of
new objects to send to the server as a thin pack.
We walk these commits until we discover a frontier of commits such
that every commit walk starting at interesting commits ends in a root
commit or unintersting commit. We then need to discover which
non-commit objects are reachable from uninteresting commits. This
commit walk is not changing during this series.
The mark_edges_uninteresting() method in list-objects.c iterates on
the commit list and does the following:
* If the commit is UNINTERSTING, then mark its root tree and every
object it can reach as UNINTERESTING.
* If the commit is interesting, then mark the root tree of every
UNINTERSTING parent (and all objects that tree can reach) as
UNINTERSTING.
At the very end, we repeat the process on every commit directly
given to the revision walk from stdin. This helps ensure we properly
cover shallow commits that otherwise were not included in the
frontier.
The logic to recursively follow trees is in the
mark_tree_uninteresting() method in revision.c. The algorithm avoids
duplicate work by not recursing into trees that are already marked
UNINTERSTING.
Add a new 'sparse' option to the mark_edges_uninteresting() method
that performs this logic in a slightly different way. As we iterate
over the commits, we add all of the root trees to an oidset. Then,
call mark_trees_uninteresting_sparse() on that oidset. Note that we
include interesting trees in this process. The current implementation
of mark_trees_unintersting_sparse() will walk the same trees as
the old logic, but this will be replaced in a later change.
Add a '--sparse' flag in 'git pack-objects' to call this new logic.
Add a new test script t/t5322-pack-objects-sparse.sh that tests this
option. The tests currently demonstrate that the resulting object
list is the same as the old algorithm. This includes a case where
both algorithms pack an object that is not needed by a remote due to
limits on the explored set of trees. When the sparse algorithm is
changed in a later commit, we will add a test that demonstrates a
change of behavior in some cases.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When parsing a tree, we read the object ID directly out of the tree
buffer. This is normally fine, but such an object ID cannot be used with
oidcpy, which copies GIT_MAX_RAWSZ bytes, because if we are using SHA-1,
there may not be that many bytes to copy.
Instead, store the object ID in a separate struct member. Since we can
no longer efficiently compute the path length, store that information as
well in struct name_entry. Ensure we only copy the object ID into the
new buffer if the path length is nonzero, as some callers will pass us
an empty path with no object ID following it, and we will not want to
read past the end of the buffer.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The traversal over tree objects has learned to honor
":(attr:label)" pathspec match, which has been implemented only for
enumerating paths on the filesystem.
* nd/attr-pathspec-in-tree-walk:
tree-walk: support :(attr) matching
dir.c: move, rename and export match_attrs()
pathspec.h: clean up "extern" in function declarations
tree-walk.c: make tree_entry_interesting() take an index
tree.c: make read_tree*() take 'struct repository *'
In order to support :(attr) when matching pathspec on a tree,
tree_entry_interesting() needs to take an index (because
git_check_attr() needs it). This is the preparation step for it. This
also makes it clearer what index we fall back to when looking up
attributes during an unpack-trees operation: the source index.
This also fixes revs->pruning.repo initialization that should have
been done in 2abf350385 (revision.c: remove implicit dependency on
the_index - 2018-09-21). Without it, skip_uninteresting() will
dereference a NULL pointer through this call chain
get_revision(revs)
get_revision_internal
get_revision_1
try_to_simplify_commit
rev_compare_tree
diff_tree_oid(..., &revs->pruning)
ll_diff_tree_oid
diff_tree_paths
ll_diff_tree
skip_uninteresting
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While at there, since we have access to struct repository now,
eliminate the only the_repository reference in this file.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "rev-list --filter" feature learned to exclude all trees via
"tree:0" filter.
* md/filter-trees:
list-objects: support for skipping tree traversal
filter-trees: code clean-up of tests
list-objects-filter: implement filter tree:0
list-objects-filter-options: do not over-strbuf_init
list-objects-filter: use BUG rather than die
revision: mark non-user-given objects instead
rev-list: handle missing tree objects properly
list-objects: always parse trees gently
list-objects: refactor to process_tree_contents
list-objects: store common func args in struct
The tree:0 filter does not need to traverse the trees that it has
filtered out, so optimize list-objects and list-objects-filter to skip
traversing the trees entirely. Before this patch, we iterated over all
children of the tree, and did nothing for all of them, which was
wasteful.
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, list-objects.c incorrectly treats all root trees of commits
as USER_GIVEN. Also, it would be easier to mark objects that are
non-user-given instead of user-given, since the places in the code
where we access an object through a reference are more obvious than
the places where we access an object that was given by the user.
Resolve these two problems by introducing a flag NOT_USER_GIVEN that
marks blobs and trees that are non-user-given, replacing USER_GIVEN.
(Only blobs and trees are marked because this mark is only used when
filtering objects, and filtering of other types of objects is not
supported yet.)
This fixes a bug in that git rev-list behaved differently from git
pack-objects. pack-objects would *not* filter objects given explicitly
on the command line and rev-list would filter. This was because the two
commands used a different function to add objects to the rev_info
struct. This seems to have been an oversight, and pack-objects has the
correct behavior, so I added a test to make sure that rev-list now
behaves properly.
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Previously, we assumed only blob objects could be missing. This patch
makes rev-list handle missing trees like missing blobs. The --missing=*
and --exclude-promisor-objects flags now work for trees as they already
do for blobs. This is demonstrated in t6112.
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If parsing fails when revs->ignore_missing_links and
revs->exclude_promisor_objects are both false, we print the OID anyway
in the die("bad tree object...") call, so any message printed by
parse_tree_gently() is superfluous.
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This will be used in a follow-up patch to reduce indentation needed when
invoking the logic conditionally. i.e. rather than:
if (foo) {
while (...) {
/* this is very indented */
}
}
we will have:
if (foo)
process_tree_contents(...);
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This will make utility functions easier to create, as done by the next
patch.
Signed-off-by: Matthew DeVore <matvore@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Partial clone support of "git clone" has been updated to correctly
validate the objects it receives from the other side. The server
side has been corrected to send objects that are directly
requested, even if they may match the filtering criteria (e.g. when
doing a "lazy blob" partial clone).
* jt/partial-clone-fsck-connectivity:
clone: check connectivity even if clone is partial
upload-pack: send refs' objects despite "filter"
The conversion to pass "the_repository" and then "a_repository"
throughout the object access API continues.
* sb/object-store-grafts:
commit: allow lookup_commit_graft to handle arbitrary repositories
commit: allow prepare_commit_graft to handle arbitrary repositories
shallow: migrate shallow information into the object parser
path.c: migrate global git_path_* to take a repository argument
cache: convert get_graft_file to handle arbitrary repositories
commit: convert read_graft_file to handle arbitrary repositories
commit: convert register_commit_graft to handle arbitrary repositories
commit: convert commit_graft_pos() to handle arbitrary repositories
shallow: add repository argument to is_repository_shallow
shallow: add repository argument to check_shallow_file_for_update
shallow: add repository argument to register_shallow
shallow: add repository argument to set_alternate_shallow_file
commit: add repository argument to lookup_commit_graft
commit: add repository argument to prepare_commit_graft
commit: add repository argument to read_graft_file
commit: add repository argument to register_commit_graft
commit: add repository argument to commit_graft_pos
object: move grafts to object parser
object-store: move object access functions to object-store.h
A filter line in a request to upload-pack filters out objects regardless
of whether they are directly referenced by a "want" line or not. This
means that cloning with "--filter=blob:none" (or another filter that
excludes blobs) from a repository with at least one ref pointing to a
blob (for example, the Git repository itself) results in output like the
following:
error: missing object referenced by 'refs/tags/junio-gpg-pub'
and if that particular blob is not referenced by a fetched tree, the
resulting clone fails fsck because there is no object from the remote to
vouch that the missing object is a promisor object.
Update both the protocol and the upload-pack implementation to include
all explicitly specified "want" objects in the packfile regardless of
the filter specification.
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a repository argument to allow the callers of lookup_tree
to be more specific about which repository to act on. This is a small
mechanical change; it doesn't change the implementation to handle
repositories other than the_repository yet.
As with the previous commits, use a macro to catch callers passing a
repository other than the_repository at compile time.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a repository argument to allow the callers of lookup_blob
to be more specific about which repository to act on. This is a small
mechanical change; it doesn't change the implementation to handle
repositories other than the_repository yet.
As with the previous commits, use a macro to catch callers passing a
repository other than the_repository at compile time.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This should make these functions easier to find and cache.h less
overwhelming to read.
In particular, this moves:
- read_object_file
- oid_object_info
- write_object_file
As a result, most of the codebase needs to #include object-store.h.
In this patch the #include is only added to files that would fail to
compile otherwise. It would be better to #include wherever
identifiers from the header are used. That can happen later
when we have better tooling for it.
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In anticipation of making trees load lazily, create a Coccinelle
script (contrib/coccinelle/commit.cocci) to ensure that all
references to the 'maybe_tree' member of struct commit are either
mutations or accesses through get_commit_tree() or
get_commit_tree_oid().
Apply the Coccinelle script to create the rest of the patch.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Using the commit-graph file to walk commit history removes the large
cost of parsing commits during the walk. This exposes a performance
issue: lookup_tree() takes a large portion of the computation time,
even when Git never uses those trees.
In anticipation of lazy-loading these trees, rename the 'tree' member
of struct commit to 'maybe_tree'. This serves two purposes: it hints
at the future role of possibly being NULL even if the commit has a
valid tree, and it allows for unambiguous transformation from simple
member access (i.e. commit->maybe_tree) to method access.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In preparation for implementing narrow/partial clone, the machinery
for checking object connectivity used by gc and fsck has been
taught that a missing object is OK when it is referenced by a
packfile specially marked as coming from trusted repository that
promises to make them available on-demand and lazily.
* jh/fsck-promisors:
gc: do not repack promisor packfiles
rev-list: support termination at promisor objects
sha1_file: support lazily fetching missing objects
introduce fetch-object: fetch one promisor object
index-pack: refactor writing of .keep files
fsck: support promisor objects as CLI argument
fsck: support referenced promisor objects
fsck: support refs pointing to promisor objects
fsck: introduce partialclone extension
extension.partialclone: introduce partial clone extension
"git describe" was taught to dig trees deeper to find a
<commit-ish>:<path> that refers to a given blob object.
* sb/describe-blob:
builtin/describe.c: describe a blob
builtin/describe.c: factor out describe_commit
builtin/describe.c: print debug statements earlier
builtin/describe.c: rename `oid` to avoid variable shadowing
revision.h: introduce blob/tree walking in order of the commits
list-objects.c: factor out traverse_trees_and_blobs
t6120: fix typo in test name
Teach rev-list to support termination of an object traversal at any
object from a promisor remote (whether one that the local repo also has,
or one that the local repo knows about because it has another promisor
object that references it).
This will be used subsequently in gc and in the connectivity check used
by fetch.
For efficiency, if an object is referenced by a promisor object, and is
in the local repo only as a non-promisor object, object traversal will
not stop there. This is to avoid building the list of promisor object
references.
(In list-objects.c, the case where obj is NULL in process_blob() and
process_tree() do not need to be changed because those happen only when
there is a conflict between the expected type and the existing object.
If the object doesn't exist, an object will be synthesized, which is
fine.)
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create traverse_commit_list_filtered() and add filtering
interface to allow certain objects to be omitted from the
traversal.
Update traverse_commit_list() to be a wrapper for the above
with a null filter to minimize the number of callers that
needed to be changed.
Object filtering will be used in a future commit by rev-list
and pack-objects for partial clone and fetch to omit unwanted
objects from the result.
traverse_bitmap_commit_list() does not work with filtering.
If a packfile bitmap is present, it will not be used. It
should be possible to extend such support in the future (at
least to simple filters that do not require object pathnames),
but that is beyond the scope of this patch series.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Reviewed-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The functionality to list tree objects in the order they were seen
while traversing the commits will be used in one of the next commits,
where we teach `git describe` to describe not only commits, but blobs, too.
The change in list-objects.c is rather minimal as we'll be re-using
the infrastructure put in place of the revision walking machinery. For
example one could expect that add_pending_tree is not called, but rather
commit->tree is directly passed to the tree traversal function. This
however requires a lot more code than just emptying the queue containing
trees after each commit.
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With traverse_trees_and_blobs factored out of the main traverse function,
the next patch can introduce an in-order revision walking with ease.
In the next patch we'll call `traverse_trees_and_blobs` from within the
loop walking the commits, such that we'll have one invocation of that
function per commit. That is why we do not want to have memory allocations
in that function, such as we'd have if we were to use a strbuf locally.
Pass a strbuf from traverse_commit_list into the blob and tree traversing
function as a scratch pad that only needs to be allocated once.
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Convert the lookup_tree function to take a pointer to struct object_id.
The commit was created with manual changes to tree.c, tree.h, and
object.c, plus the following semantic patch:
@@
@@
- lookup_tree(EMPTY_TREE_SHA1_BIN)
+ lookup_tree(&empty_tree_oid)
@@
expression E1;
@@
- lookup_tree(E1.hash)
+ lookup_tree(&E1)
@@
expression E1;
@@
- lookup_tree(E1->hash)
+ lookup_tree(E1)
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Convert lookup_blob to take a pointer to struct object_id.
The commit was created with manual changes to blob.c and blob.h, plus
the following semantic patch:
@@
expression E1;
@@
- lookup_blob(E1.hash)
+ lookup_blob(&E1)
@@
expression E1;
@@
- lookup_blob(E1->hash)
+ lookup_blob(E1)
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we find a blob at "a/b/c", we currently pass this to
our show_object_fn callbacks as two components: "a/b/" and
"c". Callbacks which want the full value then call
path_name(), which concatenates the two. But this is an
inefficient interface; the path is a strbuf, and we could
simply append "c" to it temporarily, then roll back the
length, without creating a new copy.
So we could improve this by teaching the callsites of
path_name() this trick (and there are only 3). But we can
also notice that no callback actually cares about the
broken-down representation, and simply pass each callback
the full path "a/b/c" as a string. The callback code becomes
even simpler, then, as we do not have to worry about freeing
an allocated buffer, nor rolling back our modification to
the strbuf.
This is theoretically less efficient, as some callbacks
would not bother to format the final path component. But in
practice this is not measurable. Since we use the same
strbuf over and over, our work to grow it is amortized, and
we really only pay to memcpy a few bytes.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the previous commit, we left name_path as a thin wrapper
around a strbuf. This patch drops it entirely. As a result,
every show_object_fn callback needs to be adjusted. However,
none of their code needs to be changed at all, because the
only use was to pass it to path_name(), which now handles
the bare strbuf.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "struct name_path" data is examined in only two places:
we generate it in process_tree(), and we convert it to a
single string in path_name(). Everyone else just passes it
through to those functions.
We can further note that process_tree() already keeps a
single strbuf with the leading tree path, for use with
tree_entry_interesting().
Instead of building a separate name_path linked list, let's
just use the one we already build in "base". This reduces
the amount of code (especially tricky code in path_name()
which did not check for integer overflows caused by deep
or large pathnames).
It is also more efficient in some instances. Any time we
were using tree_entry_interesting, we were building up the
strbuf anyway, so this is an immediate and obvious win
there. In cases where we were not, we trade off storing
"pathname/" in a strbuf on the heap for each level of the
path, instead of two pointers and an int on the stack (with
one pointer into the tree object). On a 64-bit system, the
latter is 20 bytes; so if path components are less than that
on average, this has lower peak memory usage. In practice
it probably doesn't matter either way; we are already
holding in memory all of the tree objects leading up to each
pathname, and for normal-depth pathnames, we are only
talking about hundreds of bytes.
This patch leaves "struct name_path" as a thin wrapper
around the strbuf, to avoid disrupting callbacks. We should
fix them, but leaving it out makes this diff easier to view.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
struct object is one of the major data structures dealing with object
IDs. Convert it to use struct object_id instead of an unsigned char
array. Convert get_object_hash to refer to the new member as well.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Jeff King <peff@peff.net>
Recent "git prune" traverses young unreachable objects to safekeep
old objects in the reachability chain from them, which sometimes
caused error messages that are unnecessarily alarming.
* jk/squelch-missing-link-warning-for-unreachable:
suppress errors on missing UNINTERESTING links
silence broken link warnings with revs->ignore_missing_links
add quieter versions of parse_{tree,commit}
We set revs->ignore_missing_links to instruct the
revision-walking machinery that we know the history graph
may be incomplete. For example, we use it when walking
unreachable but recent objects; we want to add what we can,
but it's OK if the history is incomplete.
However, we still print error messages for the missing
objects, which can be confusing. This is not an error, but
just a normal situation when transitioning from a repository
last pruned by an older git (which can leave broken segments
of history) to a more recent one (where we try to preserve
whole reachable segments).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In commit fbd4a70 (list-objects: mark more commits as edges in
mark_edges_uninteresting - 2013-08-16), we marked an increasing number
of edges uninteresting. This change, and the subsequent change to make
this conditional on --objects-edge, are used by --thin to make much
smaller packs for shallow clones.
Unfortunately, they cause a significant performance regression when
pushing non-shallow clones with lots of refs (23.322 seconds vs.
4.785 seconds with 22400 refs). Add an option to git rev-list,
--objects-edge-aggressive, that preserves this more aggressive behavior,
while leaving --objects-edge to provide more performant behavior.
Preserve the current behavior for the moment by using the aggressive
option.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we call traverse_commit_list, we may have trees and
blobs in the pending array. As we process these, we pass the
"name" field from the pending entry as the path of the
object within the tree (which then becomes the root path if
we recurse into subtrees).
When we set up the traversal in prepare_revision_walk,
though, the "name" field of any pending trees and blobs is
likely to be the ref at which we found the object. We would
not want to make this part of the path (e.g., doing so would
make "git rev-list --objects v2.6.11-tree" in linux.git show
paths like "v2.6.11-tree/Makefile", which is nonsensical).
Therefore prepare_revision_walk sets the name field of each
pending tree and blobs to the empty string.
However, this leaves no room for a caller who does know the
correct path of a pending object to propagate that
information to the revision walker. We can fix this by
making two related changes:
1. Use the "path" field as the path instead of the "name"
field in traverse_commit_list. If the path is not set,
default to "" (which is what we always ended up with in
the current code, because of prepare_revision_walk).
2. In prepare_revision_walk, make a complete copy of the
entry. This makes the path field available to the
walker (if there is one), solving our problem.
Leaving the name field intact is now OK, as we do not
use it as a path due to point (1) above (and we can use
it to make more meaningful error messages if we want).
We also make the original "mode" field available to the
walker, though it does not actually use it.
Note that we still re-add the pending objects and free the
old ones (so we may strdup the path and name only to free
the old ones). This could be made more efficient by simply
copying the object_array entries that we are keeping.
However, that would require more restructuring of the code,
and is not done here.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There's currently no easy way to free the memory associated
with an object_array (and in most cases, we simply leak the
memory in a rev_info's pending array). Let's provide a
helper to make this easier to handle.
We can make use of it in list-objects.c, which does the same
thing by hand (but fails to free the "name" field of each
entry, potentially leaking memory).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When pack-objects is computing the reachability bitmap to
serve a fetch request, it can erroneously die() if some of
the UNINTERESTING objects are not present. Upload-pack
throws away HAVE lines from the client for objects we do not
have, but we may have a tip object without all of its
ancestors (e.g., if the tip is no longer reachable and was
new enough to survive a `git prune`, but some of its
reachable objects did get pruned).
In the non-bitmap case, we do a revision walk with the HAVE
objects marked as UNINTERESTING. The revision walker
explicitly ignores errors in accessing UNINTERESTING commits
to handle this case (and we do not bother looking at
UNINTERESTING trees or blobs at all).
When we have bitmaps, however, the process is quite
different. The bitmap index for a pack-objects run is
calculated in two separate steps:
First, we perform an extensive walk from all the HAVEs to
find the full set of objects reachable from them. This walk
is usually optimized away because we are expected to hit an
object with a bitmap during the traversal, which allows us
to terminate early.
Secondly, we perform an extensive walk from all the WANTs,
which usually also terminates early because we hit a commit
with an existing bitmap.
Once we have the resulting bitmaps from the two walks, we
AND-NOT them together to obtain the resulting set of objects
we need to pack.
When we are walking the HAVE objects, the revision walker
does not know that we are walking it only to mark the
results as uninteresting. We strip out the UNINTERESTING flag,
because those objects _are_ interesting to us during the
first walk. We want to keep going to get a complete set of
reachable objects if we can.
We need some way to tell the revision walker that it's OK to
silently truncate the HAVE walk, just like it does for the
UNINTERESTING case. This patch introduces a new
`ignore_missing_links` flag to the `rev_info` struct, which
we set only for the HAVE walk.
It also adds tests to cover UNINTERESTING objects missing
from several positions: a missing blob, a missing tree, and
a missing parent commit. The missing blob already worked (as
we do not care about its contents at all), but the other two
cases caused us to die().
Note that there are a few cases we do not need to test:
1. We do not need to test a missing tree, with the blob
still present. Without the tree that refers to it, we
would not know that the blob is relevant to our walk.
2. We do not need to test a tip commit that is missing.
Upload-pack omits these for us (and in fact, we
complain even in the non-bitmap case if it fails to do
so).
Reported-by: Siddharth Agarwal <sid0@fb.com>
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix performance regression in v1.8.4.x and later.
* jk/mark-edges-uninteresting:
list-objects: only look at cmdline trees with edge_hint
t/perf: time rev-list with UNINTERESTING commits
When rev-list is given a command-line like:
git rev-list --objects $commit --not --all
the most accurate answer is the difference between the set
of objects reachable from $commit and the set reachable from
all of the existing refs. However, we have not historically
provided that answer, because it is very expensive to
calculate. We would have to open every tree of every commit
in the entire history.
Instead, we find the accurate set difference of the
reachable commits, and then mark the trees at the boundaries
as uninteresting. This misses objects which appear in the
trees of both the interesting commits and deep within the
uninteresting history.
Commit fbd4a70 (list-objects: mark more commits as edges in
mark_edges_uninteresting, 2013-08-16) noticed that we miss
those objects during pack-objects, and added code to examine
the trees of all of the "--not" refs given on the
command-line. Note that this is still not the complete set
difference, because we look only at the tips of the
command-line arguments, not all of their reachable commits.
But it increases the set of boundary objects we consider,
which is especially important for shallow fetches. So we
are trading extra CPU time for a larger set of boundary
objects, which can improve the resulting pack size for a
--thin pack.
This tradeoff probably makes sense in the context of
pack-objects, where we have set revs->edge_hint to have the
traversal feed us the set of boundary objects. For a
regular rev-list, though, it is probably not a good
tradeoff. It is true that it makes our list slightly closer
to a true set difference, but it is a rare case where this
is important. And because we do not have revs->edge_hint
set, we do nothing useful with the larger set of boundary
objects.
This patch therefore ties the extra tree examination to the
revs->edge_hint flag; it is the presence of that flag that
makes the tradeoff worthwhile.
Here is output from the p0001-rev-list showing the
improvement in performance:
Test HEAD^ HEAD
-----------------------------------------------------------------------------------------
0001.1: rev-list --all 0.69(0.65+0.02) 0.69(0.66+0.02) +0.0%
0001.2: rev-list --all --objects 3.22(3.19+0.03) 3.23(3.20+0.03) +0.3%
0001.4: rev-list $commit --not --all 0.04(0.04+0.00) 0.04(0.04+0.00) +0.0%
0001.5: rev-list --objects $commit --not --all 0.27(0.26+0.01) 0.04(0.04+0.00) -85.2%
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When there is no sufficient overlap between old and new history
during a fetch into a shallow repository, we unnecessarily sent
objects the sending side knows the receiving end has.
* nd/fetch-into-shallow:
Add testcase for needless objects during a shallow fetch
list-objects: mark more commits as edges in mark_edges_uninteresting
list-objects: reduce one argument in mark_edges_uninteresting
upload-pack: delegate rev walking in shallow fetch to pack-objects
shallow: add setup_temporary_shallow()
shallow: only add shallow graft points to new shallow file
move setup_alternate_shallow and write_shallow_commits to shallow.c
The purpose of edge commits is to let pack-objects know what objects
it can use as base, but does not need to include in the thin pack
because the other side is supposed to already have them. So far we
mark uninteresting parents of interesting commits as edges. But even
an unrelated uninteresting commit (that the other side has) may
become a good base for pack-objects and help produce more efficient
packs.
This is especially true for shallow clone, when the client issues a
fetch with a depth smaller or equal to the number of commits the
server is ahead of the client. For example, in this commit history
the client has up to "A" and the server has up to "B":
-------A---B
have--^ ^
/
want--+
If depth 1 is requested, the commit list to send to the client
includes only B. The way m_e_u is working, it checks if parent
commits of B are uninteresting, if so mark them as edges. Due to
shallow effect, commit B is grafted to have no parents and the
revision walker never sees A as the parent of B. In fact it marks no
edges at all in this simple case and sends everything B has to the
client even if it could have excluded what A and also the client
already have.
In a slightly different case where A is not a direct parent of B
(iow there are commits in between A and B), marking A as an edge can
still save some because B may still have stuff from the far ancestor
A.
There is another case from the earlier patch, when we deepen a ref
from C->E to A->E:
---A---B C---D---E
want--^ ^ ^
shallow-+ /
have-------+
In this case we need to send A and B to the client, and C (i.e. the
current shallow point that the client informs the server) is a very
good base because it's closet to A and B. Normal m_e_u won't recognize
C as an edge because it only looks back to parents (i.e. A<-B) not the
opposite way B->C even if C is already marked as uninteresting commit
by the previous patch.
This patch includes all uninteresting commits from command line as
edges and lets pack-objects decide what's best to do. The upside is we
have better chance of producing better packs in certain cases. The
downside is we may need to process some extra objects on the server
side.
For the shallow case on git.git, when the client is 5 commits behind
and does "fetch --depth=3", the result pack is 99.26 KiB instead of
4.92 MiB.
Reported-and-analyzed-by: Matthijs Kooijman <matthijs@stdin.nl>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Junio C Hamano
Derrick Stolee
brian m. carlson
Nguyễn Thái Ngọc Duy
Matthew DeVore
Jonathan Tan
Stefan Beller
Jeff Hostetler
Jeff King
Vicent Marti