A consumer of the sequencer that wishes to take advantage of either the
`keep_redundant_commits` or `drop_redundant_commits` feature must also
specify `allow_empty`. However, these refer to two distinct types of
empty commits:
- `allow_empty` refers specifically to commits which start empty
- `keep_redundant_commits` refers specifically to commits that do not
start empty, but become empty due to the content already existing in
the target history
Conceptually, there is no reason that the behavior for handling one of
these should be entangled with the other. It is particularly unintuitive
to require `allow_empty` in order for `drop_redundant_commits` to have
an effect: in order to prevent redundant commits automatically,
initially-empty commits would need to be kept automatically as well.
Instead, rewrite the `allow_empty()` logic to remove the over-arching
requirement that `allow_empty` be specified in order to reach any of the
keep/drop behaviors. Only if the commit was originally empty will
`allow_empty` have an effect.
Note that no behavioral changes should result from this commit -- it
merely sets the stage for future commits. In one such future commit, an
`--empty` option will be added to git-cherry-pick(1), meaning that
`drop_redundant_commits` will be used by that command.
Signed-off-by: Brian Lyles <brianmlyles@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using git-cherry-pick(1) with `--allow-empty` while on an unborn
branch, an error is thrown. This is inconsistent with the same
cherry-pick when `--allow-empty` is not specified.
Detect unborn branches in `is_index_unchanged`. When on an unborn
branch, use the `empty_tree` as the tree to compare against.
Add a new test to cover this scenario. While modelled off of the
existing 'cherry-pick on unborn branch' test, some improvements can be
made:
- Use `git switch --orphan unborn` instead of `git checkout --orphan
unborn` to avoid the need for a separate `rm -rf *` call
- Avoid using `--quiet` in the `git diff` call to make debugging easier
in the event of a failure. Use simply `--exit-code` instead.
Make these improvements to the existing test as well as the new test.
Helped-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Brian Lyles <brianmlyles@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When git-am(1) got its own `--empty` option in 7c096b8d61 (am: support
--empty=<option> to handle empty patches, 2021-12-09), `stop` was used
instead of `ask`. `stop` is a more accurate term for describing what
really happens, and consistency is good.
Update git-rebase(1) to also use `stop`, while keeping `ask` as a
deprecated synonym. Update the tests to primarily use `stop`, but also
ensure that `ask` is still allowed.
In a future commit, we'll be adding a new `--empty` option for
git-cherry-pick(1) as well, making the consistency even more relevant.
Reported-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Brian Lyles <brianmlyles@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Both of these pages document very similar `--empty` options, but with
different styles. The exact behavior of these `--empty` options differs
somewhat, but consistent styling in the docs is still beneficial. This
commit aims to make them more consistent.
Break the possible values for `--empty` into separate sections for
readability. Alphabetical order is chosen for consistency.
In a future commit, we'll be documenting a new `--empty` option for
git-cherry-pick(1), making the consistency even more relevant.
Signed-off-by: Brian Lyles <brianmlyles@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The documentation for git-rebase(1) indicates that using the `--exec`
option will use `--empty=drop`. This is inaccurate: when `--interactive`
is not explicitly provided, `--exec` results in `--empty=keep`
behaviors.
Correctly indicate the behavior of `--exec` using `--empty=keep` when
`--interactive` is not specified.
Reported-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Brian Lyles <brianmlyles@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The upload-pack program, when talking over v2, accepted the
packfile-uris protocol extension from the client, even if it did
not advertise the capability, which has been corrected.
* jk/upload-pack-v2-capability-cleanup:
upload-pack: only accept packfile-uris if we advertised it
upload-pack: use existing config mechanism for advertisement
upload-pack: centralize setup of sideband-all config
upload-pack: use repository struct to get config
Various parts of upload-pack has been updated to bound the resource
consumption relative to the size of the repository to protect from
abusive clients.
* jk/upload-pack-bounded-resources:
upload-pack: free tree buffers after parsing
upload-pack: use PARSE_OBJECT_SKIP_HASH_CHECK in more places
upload-pack: always turn off save_commit_buffer
upload-pack: disallow object-info capability by default
upload-pack: accept only a single packfile-uri line
upload-pack: use a strmap for want-ref lines
upload-pack: use oidset for deepen_not list
upload-pack: switch deepen-not list to an oid_array
upload-pack: drop separate v2 "haves" array
Clear the fallout from a fix for 2.44 regression.
* ps/reftable-repo-init-fix:
t0610: remove unused variable assignment
refs/reftable: don't fail empty transactions in repo without HEAD
A custom remote helper no longer cannot access the newly created
repository during "git clone", which is a regression in Git 2.44.
This has been corrected.
* ps/remote-helper-repo-initialization-fix:
builtin/clone: allow remote helpers to detect repo
"git log --merge" learned to pay attention to CHERRY_PICK_HEAD and
other kinds of *_HEAD pseudorefs.
* ml/log-merge-with-cherry-pick-and-other-pseudo-heads:
revision: implement `git log --merge` also for rebase/cherry-pick/revert
revision: ensure MERGE_HEAD is a ref in prepare_show_merge
"git commit -v --cleanup=scissors" used to add the scissors line
twice in the log message buffer, which has been corrected.
* jt/commit-redundant-scissors-fix:
commit: unify logic to avoid multiple scissors lines when merging
commit: avoid redundant scissor line with --cleanup=scissors -v
"git merge-tree" has learned that the three trees involved in the
3-way merge only need to be trees, not necessarily commits.
* js/merge-tree-3-trees:
fill_tree_descriptor(): mark error message for translation
cache-tree: avoid an unnecessary check
Always check `parse_tree*()`'s return value
t4301: verify that merge-tree fails on missing blob objects
merge-ort: do check `parse_tree()`'s return value
merge-tree: fail with a non-zero exit code on missing tree objects
merge-tree: accept 3 trees as arguments
"git rev-list --missing=print" has learned to optionally take
"--allow-missing-tips", which allows the objects at the starting
points to be missing.
* cc/rev-list-allow-missing-tips:
revision: fix --missing=[print|allow*] for annotated tags
rev-list: allow missing tips with --missing=[print|allow*]
t6022: fix 'test' style and 'even though' typo
oidset: refactor oidset_insert_from_set()
revision: clarify a 'return NULL' in get_reference()
"git --no-lazy-fetch cmd" allows to run "cmd" while disabling lazy
fetching of objects from the promisor remote, which may be handy
for debugging.
* jc/no-lazy-fetch:
git: extend --no-lazy-fetch to work across subprocesses
git: document GIT_NO_REPLACE_OBJECTS environment variable
git: --no-lazy-fetch option
In b0f6b6b523 (refs/reftable: don't fail empty transactions in repo
without HEAD, 2024-02-27), we have added a new test to t0610. This test
contains a useless assignment to a variable that is never actually used.
Remove it.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Also update the DEF_VER in GIT-VERSION-GEN, which I forgot to do
earlier (it should have been done when we started the new cycle).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git for-each-ref" learned "--include-root-refs" option to show
even the stuff outside the 'refs/' hierarchy.
* kn/for-all-refs:
for-each-ref: add new option to include root refs
ref-filter: rename 'FILTER_REFS_ALL' to 'FILTER_REFS_REGULAR'
refs: introduce `refs_for_each_include_root_refs()`
refs: extract out `loose_fill_ref_dir_regular_file()`
refs: introduce `is_pseudoref()` and `is_headref()`
When a merge conflicted at a submodule, merge-ort backend used to
unconditionally give a lengthy message to suggest how to resolve
it. Now the message can be squelched as an advice message.
* pb/ort-make-submodule-conflict-message-an-advice:
merge-ort: turn submodule conflict suggestions into an advice
Clarify wording in the CodingGuidelines that requires <git-compat-util.h>
to be the first header file.
* jc/doc-compat-util:
doc: clarify the wording on <git-compat-util.h> requirement
An error message from "git upload-pack", which responds to "git
fetch" requests, had a trialing NUL in it, which has been
corrected.
* sg/upload-pack-error-message-fix:
upload-pack: don't send null character in abort message to the client
Many small allocations "git name-rev" makes have been updated to
allocate from a mem-pool.
* rs/name-rev-with-mempool:
name-rev: use mem_pool_strfmt()
mem-pool: add mem_pool_strfmt()
The logic to access reflog entries by date and number had ugly
corner cases at the boundaries, which have been cleaned up.
* jk/reflog-special-cases-fix:
read_ref_at(): special-case ref@{0} for an empty reflog
get_oid_basic(): special-case ref@{n} for oldest reflog entry
Revert "refs: allow @{n} to work with n-sized reflog"
The code incorrectly attempted to use textconv cache when asked,
even when we are not running in a repository, which has been
corrected.
* jk/textconv-cache-outside-repo-fix:
userdiff: skip textconv caching when not in a repository
Practice the new mark-up rule for <placeholders> with "git add"
documentation page.
* jc/doc-add-placeholder-fix:
doc: apply the new placeholder rules to git-add documentation
The way placeholders are to be marked-up in documentation have been
specified; use "_<placeholder>_" to typeset the word inside a pair
of <angle-brakets> emphasized.
* ja/doc-placeholders-markup-rules:
doc: clarify the format of placeholders
"git reflog" learned a "list" subcommand that enumerates known reflogs.
* ps/reflog-list:
builtin/reflog: introduce subcommand to list reflogs
refs: stop resolving ref corresponding to reflogs
refs: drop unused params from the reflog iterator callback
refs: always treat iterators as ordered
refs/files: sort merged worktree and common reflogs
refs/files: sort reflogs returned by the reflog iterator
dir-iterator: support iteration in sorted order
dir-iterator: pass name to `prepare_next_entry_data()` directly
Doc update.
* ja/docfixes:
doc: end sentences with full-stop
doc: close unclosed angle-bracket of a placeholder in git-clone doc
doc: git-rev-parse: enforce command-line description syntax
"git difftool --dir-diff" learned to honor the "--trust-exit-code"
option; it used to always exit with 0 and signalled success.
* ps/difftool-dir-diff-exit-code:
git-difftool--helper: honor `--trust-exit-code` with `--dir-diff`
The 'refresh' function in 'builtin/add.c' declares 'flags' as
signed, and passes it as an argument to the 'refresh_index'
function, which though expects an unsigned value.
Since in this case 'flags' represents a bag of bits, whose MSB is
not used in special ways, change the type of 'flags' to unsigned.
Signed-off-by: Eugenio Gigante <giganteeugenio2@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Clients are only supposed to request particular capabilities or features
if the server advertised them. For the "packfile-uris" feature, we only
advertise it if uploadpack.blobpacfileuri is set, but we always accept a
request from the client regardless.
In practice this doesn't really hurt anything, as we'd pass the client's
protocol list on to pack-objects, which ends up ignoring it. But we
should try to follow the protocol spec, and tightening this up may catch
buggy or misbehaving clients more easily.
Thanks to recent refactoring, we can hoist the config check from
upload_pack_advertise() into upload_pack_config(). Note the subtle
handling of a value-less bool (which does not count for triggering an
advertisement).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When serving a v2 capabilities request, we call upload_pack_advertise()
to tell us the set of features we can advertise to the client. That
involves looking at various config options, all of which need to be kept
in sync with the rules we use in upload_pack_config to set flags like
allow_filter, allow_sideband_all, and so on. If these two pieces of code
get out of sync then we may refuse to respect a capability we
advertised, or vice versa accept one that we should not.
Instead, let's call the same config helper that we'll use for processing
the actual client request, and then just pick the values out of the
resulting struct. This is only a little bit shorter than the current
code, but we don't repeat any policy logic (e.g., we don't have to worry
about the magic sideband-all environment variable here anymore).
And this reveals a gap in the existing code: there is no struct flag for
the packfile-uris capability (we accept it even if it is not advertised,
which we should not). We'll leave the advertisement code for now and
deal with it in the next patch.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We read uploadpack.allowsidebandall to set a matching flag in our
upload_pack_data struct. But for our tests, we also respect
GIT_TEST_SIDEBAND_ALL from the environment, and anybody looking at the
flag in the struct needs to remember to check both. There's only one
such piece of code now, but we're about to add another.
So let's have the config step actually fold the environment value into
the struct, letting the rest of the code use the flag in the obvious
way.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our upload_pack_v2() function gets a repository struct, but we ignore it
totally. In practice this doesn't cause any problems, as it will never
differ from the_repository. But in the spirit of taking a small step
towards getting rid of the_repository, let's at least starting using it
to grab config. There are probably other spots that could benefit, but
it's a start.
Note that we don't need to pass the repo for protected_config(); the
whole point there is that we are not looking at repo config, so there is
no repo-specific version of the function.
For the v0 version of the protocol, we're not passed a repository
struct, so we'll continue to use the_repository there.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a client sends us a "want" or "have" line, we call parse_object()
to get an object struct. If the object is a tree, then the parsed state
means that tree->buffer points to the uncompressed contents of the tree.
But we don't really care about it. We only really need to parse commits
and tags; for trees and blobs, the important output is just a "struct
object" with the correct type.
But much worse, we do not ever free that tree buffer. It's not leaked in
the traditional sense, in that we still have a pointer to it from the
global object hash. But if the client requests many trees, we'll hold
all of their contents in memory at the same time.
Nobody really noticed because it's rare for clients to directly request
a tree. It might happen for a lightweight tag pointing straight at a
tree, or it might happen for a "tree:depth" partial clone filling in
missing trees.
But it's also possible for a malicious client to request a lot of trees,
causing upload-pack's memory to balloon. For example, without this
patch, requesting every tree in git.git like:
pktline() {
local msg="$*"
printf "%04x%s\n" $((1+4+${#msg})) "$msg"
}
want_trees() {
pktline command=fetch
printf 0001
git cat-file --batch-all-objects --batch-check='%(objectname) %(objecttype)' |
while read oid type; do
test "$type" = "tree" || continue
pktline want $oid
done
pktline done
printf 0000
}
want_trees | GIT_PROTOCOL=version=2 valgrind --tool=massif ./git upload-pack . >/dev/null
shows a peak heap usage of ~3.7GB. Which is just about the sum of the
sizes of all of the uncompressed trees. For linux.git, it's closer to
17GB.
So the obvious thing to do is to call free_tree_buffer() after we
realize that we've parsed a tree. We know that upload-pack won't need it
later. But let's push the logic into parse_object_with_flags(), telling
it to discard the tree buffer immediately. There are two reasons for
this. One, all of the relevant call-sites already call the with_options
variant to pass the SKIP_HASH flag. So it actually ends up as less code
than manually free-ing in each spot. And two, it enables an extra
optimization that I'll discuss below.
I've touched all of the sites that currently use SKIP_HASH in
upload-pack. That drops the peak heap of the upload-pack invocation
above from 3.7GB to ~24MB.
I've also modified the caller in get_reference(); a partial clone
benefits from its use in pack-objects for the reasons given in
0bc2557951 (upload-pack: skip parse-object re-hashing of "want" objects,
2022-09-06), where we were measuring blob requests. But note that the
results of get_reference() are used for traversing, as well; so we
really would _eventually_ use the tree contents. That makes this at
first glance a space/time tradeoff: we won't hold all of the trees in
memory at once, but we'll have to reload them each when it comes time to
traverse.
And here's where our extra optimization comes in. If the caller is not
going to immediately look at the tree contents, and it doesn't care
about checking the hash, then parse_object() can simply skip loading the
tree entirely, just like we do for blobs! And now it's not a space/time
tradeoff in get_reference() anymore. It's just a lazy-load: we're
delaying reading the tree contents until it's time to actually traverse
them one by one.
And of course for upload-pack, this optimization means we never load the
trees at all, saving lots of CPU time. Timing the "every tree from
git.git" request above shows upload-pack dropping from 32 seconds of CPU
to 19 (the remainder is mostly due to pack-objects actually sending the
pack; timing just the upload-pack portion shows we go from 13s to
~0.28s).
These are all highly gamed numbers, of course. For real-world
partial-clone requests we're saving only a small bit of time in
practice. But it does help harden upload-pack against malicious
denial-of-service attacks.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In commit 0bc2557951 (upload-pack: skip parse-object re-hashing of
"want" objects, 2022-09-06), we optimized the parse_object() calls for
v2 "want" lines from the client so that they avoided parsing blobs, and
so that they used the commit-graph rather than parsing commit objects
from scratch.
We should extend that to two other spots:
1. We parse "have" objects in the got_oid() function. These won't
generally be non-commits (unlike "want" lines from a partial
clone). But we still benefit from the use of the commit-graph.
2. For v0, the "want" lines are parsed in receive_needs(). These are
also less likely to be non-commits because by default they have to
be ref tips. There are config options you might set to allow
non-tip objects, but you'd mostly do so to support partial clones,
and clients recent enough to support partial clone will generally
speak v2 anyway.
So I don't expect this change to improve performance much for day-to-day
operations. But both are possible denial-of-service vectors, where an
attacker can waste our time by sending over a large number of objects to
parse (of course we may waste even more time serving a pack to them, but
we try as much as possible to optimize that in pack-objects; we should
do what we can here in upload-pack, too).
With this patch, running p5600 with GIT_TEST_PROTOCOL_VERSION=0 shows
similar results to what we saw in 0bc2557951 (which ran with the v2
protocol by default). Here are the numbers for linux.git:
Test HEAD^ HEAD
-----------------------------------------------------------------------------
5600.3: checkout of result 50.91(87.95+2.93) 41.75(79.00+3.18) -18.0%
Or for a more extreme (and malicious) case, we can claim to "have" every
blob in git.git over the v0 protocol:
$ {
echo "0032want $(git rev-parse HEAD)"
printf 0000
git cat-file --batch-all-objects --batch-check='%(objectname) %(objecttype)' |
perl -alne 'print "0032have $F[0]" if $F[1] eq "blob"'
} >input
$ time ./git.old upload-pack . <input >/dev/null
real 0m52.951s
user 0m51.633s
sys 0m1.304s
$ time ./git.new upload-pack . <input >/dev/null
real 0m0.261s
user 0m0.156s
sys 0m0.105s
(Note that these don't actually compute a pack because of the hacky
protocol usage, so those numbers are representing the raw blob-parsing
effort done by upload-pack).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the client sends us "want $oid" lines, we call parse_object($oid)
to get an object struct. It's important to parse the commits because we
need to traverse them in the negotiation phase. But of course we don't
need to hold on to the commit messages for each one.
We've turned off the save_commit_buffer flag in get_common_commits() for
a long time, since f0243f26f6 (git-upload-pack: More efficient usage of
the has_sha1 array, 2005-10-28). That helps with the commits we see
while actually traversing. But:
1. That function is only used by the v0 protocol. I think the v2
protocol's code path leaves the flag on (and thus pays the extra
memory penalty), though I didn't measure it specifically.
2. If the client sends us a bunch of "want" lines, that happens before
the negotiation phase. So we'll hold on to all of those commit
messages. Generally the number of "want" lines scales with the
refs, not with the number of objects in the repo. But a malicious
client could send a lot in order to waste memory.
As an example of (2), if I generate a request to fetch all commits in
git.git like this:
pktline() {
local msg="$*"
printf "%04x%s\n" $((1+4+${#msg})) "$msg"
}
want_commits() {
pktline command=fetch
printf 0001
git cat-file --batch-all-objects --batch-check='%(objectname) %(objecttype)' |
while read oid type; do
test "$type" = "commit" || continue
pktline want $oid
done
pktline done
printf 0000
}
want_commits | GIT_PROTOCOL=version=2 valgrind --tool=massif git-upload-pack . >/dev/null
before this patch upload-pack peaks at ~125MB, and after at ~35MB. The
difference is not coincidentally about the same as the sum of all commit
object sizes as computed by:
git cat-file --batch-all-objects --batch-check='%(objecttype) %(objectsize)' |
perl -alne '$v += $F[1] if $F[0] eq "commit"; END { print $v }'
In a larger repository like linux.git, that number is ~1GB.
In a repository with a full commit-graph file this will have no impact
(and the commit graph would save us from parsing at all, so is a much
better solution!). But it's easy to do, might help a little in
real-world cases (where even if you have a commit graph it might not be
fully up to date), and helps a lot for a worst-case malicious request.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We added an "object-info" capability to the v2 upload-pack protocol in
a2ba162cda (object-info: support for retrieving object info,
2021-04-20). In the almost 3 years since, we have not added any
client-side support, and it does not appear to exist in other
implementations either (JGit understands the verb on the server side,
but not on the client side).
Since this largely unused code is accessible over the network by
default, it increases the attack surface of upload-pack. I don't know of
any particularly severe problem, but one issue is that because of the
request/response nature of the v2 protocol, it will happily read an
unbounded number of packets, adding each one to a string list (without
regard to whether they are objects we know about, duplicates, etc).
This may be something we want to improve in the long run, but in the
short term it makes sense to disable the feature entirely. We'll add a
config option as an escape hatch for anybody who wants to develop the
feature further.
A more gentle option would be to add the config option to let people
disable it manually, but leave it enabled by default. But given that
there's no client side support, that seems like the wrong balance with
security.
Disabling by default will slow adoption a bit once client-side support
does become available (there were some patches[1] in 2022, but nothing
got merged and there's been nothing since). But clients have to deal
with older servers that do not understand the option anyway (and the
capability system handles that), so it will just be a matter of servers
flipping their config at that point (and hopefully once any unbounded
allocations have been addressed).
[jk: this is a patch that GitHub has been running for several years, but
rebased forward and with a new commit message for upstream]
[1] https://lore.kernel.org/git/20220208231911.725273-1-calvinwan@google.com/
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Brian Lyles
Junio C Hamano
Patrick Steinhardt
Eugenio Gigante
Jeff King
Taylor Blau