When we run `branch --merged`, we use prepare_revision_walk
with the merge-filter marked as UNINTERESTING. Any branch
tips that are marked UNINTERESTING after it returns must be
ancestors of that commit. As we iterate through the list of
refs to show, we check item->commit->object.flags to see
whether it was marked.
This interacts badly with --verbose, which will do a
separate walk to find the ahead/behind information for each
branch. There are two bad things that can happen:
1. The ahead/behind walk may get the wrong results,
because it can see a bogus UNINTERESTING flag leftover
from the merge-filter walk.
2. We may omit some branches if their tips are involved in
the ahead/behind traversal of a branch shown earlier.
The ahead/behind walk carefully cleans up its commit
flags, meaning it may also erase the UNINTERESTING
flag that we expect to check later.
We can solve this by moving the merge-filter state for each
ref into its "struct ref_item" as soon as we finish the
merge-filter walk. That fixes (2). Then we are free to clear
the commit flags we used in the walk, fixing (1).
Note that we actually do away with the matches_merge_filter
helper entirely here, and inline it between the revision
walk and the flag-clearing. This ensures that nobody
accidentally calls it at the wrong time (it is only safe to
check in that instant between the setting and clearing of
the global flag).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The macro ALLOC_GROW manages several aspects of dynamic memory
allocations for arrays: It performs overprovisioning in order to avoid
reallocations in future calls, updates the allocation size variable,
multiplies the item size and thus allows users to simply specify the
item count, performs the reallocation and updates the array pointer.
Sometimes this is too much. Add the macro REALLOC_ARRAY, which only
takes care of the latter three points and allows users to specfiy the
number of items the array can store. It can increase and also decrease
the size. Using the macro avoid duplicating the variable name and
takes care of the item sizes automatically.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When operating with the stateless RPC mode, we will receive a nonce
issued by another instance of us that advertised our capability and
refs some time ago. Update the logic to check received nonce to
detect this case, compute how much time has passed since the nonce
was issued and report the status with a new environment variable
GIT_PUSH_CERT_NONCE_SLOP to the hooks.
GIT_PUSH_CERT_NONCE_STATUS will report "SLOP" in such a case. The
hooks are free to decide how large a slop it is willing to accept.
Strictly speaking, the "nonce" is not really a "nonce" anymore in
the stateless RPC mode, as it will happily take any "nonce" issued
by it (which is protected by HMAC and its secret key) as long as it
is fresh enough. The degree of this security degradation, relative
to the native protocol, is about the same as the "we make sure that
the 'git push' decided to update our refs with new objects based on
the freshest observation of our refs by making sure the values they
claim the original value of the refs they ask us to update exactly
match the current state" security is loosened to accomodate the
stateless RPC mode in the existing code without this series, so
there is no need for those who are already using smart HTTP to push
to their repositories to be alarmed any more than they already are.
In addition, the server operator can set receive.certnonceslop
configuration variable to specify how stale a nonce can be (in
seconds). When this variable is set, and if the nonce received in
the certificate that passes the HMAC check was less than that many
seconds old, hooks are given "OK" in GIT_PUSH_CERT_NONCE_STATUS
(instead of "SLOP") and the received nonce value is given in
GIT_PUSH_CERT_NONCE, which makes it easier for a simple-minded
hook to check if the certificate we received is recent enough.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "--signed" option received by "git push" is first passed to the
transport layer, which the native transport directly uses to notice
that a push certificate needs to be sent. When the transport-helper
is involved, however, the option needs to be told to the helper with
set_helper_option(), and the helper needs to take necessary action.
For the smart-HTTP helper, the "necessary action" involves spawning
the "git send-pack" subprocess with the "--signed" option.
Once the above all gets wired in, the smart-HTTP transport now can
use the push certificate mechanism to authenticate its pushes.
Add a test that is modeled after tests for the native transport in
t5534-push-signed.sh to t5541-http-push-smart.sh. Update the test
Apache configuration to pass GNUPGHOME environment variable through.
As PassEnv would trigger warnings for an environment variable that
is not set, export it from test-lib.sh set to a harmless value when
GnuPG is not being used in the tests.
Note that the added test is deliberately loose and does not check
the nonce in this step. This is because the stateless RPC mode is
inevitably flaky and a nonce that comes back in the actual push
processing is one issued by a different process; if the two
interactions with the server crossed a second boundary, the nonces
will not match and such a check will fail. A later patch in the
series will work around this shortcoming.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In order to prevent a valid push certificate for pushing into an
repository from getting replayed in a different push operation, send
a nonce string from the receive-pack process and have the signer
include it in the push certificate. The receiving end uses an HMAC
hash of the path to the repository it serves and the current time
stamp, hashed with a secret seed (the secret seed does not have to
be per-repository but can be defined in /etc/gitconfig) to generate
the nonce, in order to ensure that a random third party cannot forge
a nonce that looks like it originated from it.
The original nonce is exported as GIT_PUSH_CERT_NONCE for the hooks
to examine and match against the value on the "nonce" header in the
certificate to notice a replay, but returned "nonce" header in the
push certificate is examined by receive-pack and the result is
exported as GIT_PUSH_CERT_NONCE_STATUS, whose value would be "OK"
if the nonce recorded in the certificate matches what we expect, so
that the hooks can more easily check.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The pre-receive and post-receive hooks were designed to be an
improvement over old style update and post-update hooks, which take
the update information on their command line and are limited by the
command line length limit. The same information is fed from the
standard input to pre/post-receive hooks instead to lift this
limitation. It has been mandatory for these new style hooks to
consume the update information fully from the standard input stream.
Otherwise, they would risk killing the receive-pack process via
SIGPIPE.
If a hook does not want to look at all the information, it is easy
to send its standard input to /dev/null (perhaps a niche use of hook
might need to know only the fact that a push was made, without
having to know what objects have been pushed to update which refs),
and this has already been done by existing hooks that are written
carefully.
However, because there is no good way to consistently fail hooks
that do not consume the input fully (a small push may result in a
short update record that may fit within the pipe buffer, to which
the receive-pack process may manage to write before the hook has a
chance to exit without reading anything, which will not result in a
death-by-SIGPIPE of receive-pack), it can lead to a hard to diagnose
"once in a blue moon" phantom failure.
Lift this "hooks must consume their input fully" mandate. A mandate
that is not enforced strictly is not helping us to catch mistakes in
hooks. If a hook has a good reason to decide the outcome of its
operation without reading the information we feed it, let it do so
as it pleases.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Current syntax description makes one wonder if there is any
syntactic way to distinguish between <branch> and <upstream> so that
one can specify <branch> but not <upstream>, but that is not the
case.
Make it explicit that these arguments are positional, i.e. the
earlier ones cannot be omitted if you want to give later ones.
Signed-off-by: Sergey Organov <sorganov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use `test_must_be_be_empty <file>` instead of `test -z "$(cat <file>)"`.
Suggested-by: Fabian Ruch <bafain@gmail.com>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the stderr of "git credential-cache" is redirected to a
pipe, the reader on the other end of a pipe may be surprised
that the pipe remains open long after the process exits.
This happens because we may auto-spawn a daemon which is
long-lived, and which keeps stderr open.
We can solve this by redirecting the daemon's stderr to
/dev/null once we are ready to go into our event loop. We
would not want to do so before then, because we may want to
report errors about the setup (e.g., failure to establish
the listening socket).
This does mean that we will not report errors we encounter
for specific clients. That's acceptable, as such errors
should be rare (e.g., clients sending buggy requests).
However, we also provide an escape hatch: if you want to see
these later messages, you can provide the "--debug" option
to keep stderr open.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commit 81c5cf7 (mailinfo: skip bogus UNIX From line inside
body, 2006-05-21), we have treated lines like ">From" in the body as
headers. This makes "git am" work for people who erroneously paste
the whole output from format-patch:
From 12345abcd...fedcba543210 Mon Sep 17 00:00:00 2001
From: them
Subject: [PATCH] whatever
into their email body (assuming that an mbox writer then quotes
"From" as ">From", as otherwise we would actually mailsplit on the
in-body line).
However, this has false positives if somebody actually has a commit
body that starts with "From "; in this case we erroneously remove
the line entirely from the commit message. We can make this check
more robust by making sure the line actually looks like a real mbox
"From" line.
Inspect the line that begins with ">From " a more carefully to only
skip lines that match the expected pattern (note that the datestamp
part of the format-patch output is designed to be kept constant to
help those who write magic(5) entries).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Record the URL of the intended recipient for a push (after
anonymizing it if it has authentication material) on a new "pushee
URL" header. Because the networking configuration (SSH-tunnels,
proxies, etc.) on the pushing user's side varies, the receiving
repository may not know the single canonical URL all the pushing
users would refer it as (besides, many sites allow pushing over
ssh://host/path and https://host/path protocols to the same
repository but with different local part of the path). So this
value may not be reliably used for replay-attack prevention
purposes, but this will still serve as a human readable hint to
identify the repository the certificate refers to.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With the interim protocol, we used to send the update commands even
though we already send a signed copy of the same information when
push certificate is in use. Update the send-pack/receive-pack pair
not to do so.
The notable thing on the receive-pack side is that it makes sure
that there is no command sent over the traditional protocol packet
outside the push certificate. Otherwise a pusher can claim to be
pushing one set of ref updates in the signed certificate while
issuing commands to update unrelated refs, and such an update will
evade later audits.
Finally, start documenting the protocol.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We would want to update the interim protocol so that we do not send
the usual update commands when the push certificate feature is in
use, as the same information is in the certificate. Once that
happens, the push-cert packet may become the only protocol command,
but then there is no packet to put the feature request behind, like
we always did.
As we have prepared the receiving end that understands the push-cert
feature to accept the feature request on the first protocol packet
(other than "shallow ", which was an unfortunate historical mistake
that has to come before everything else), we can give the feature
request on the push-cert packet instead of the first update protocol
packet, in preparation for the next step to actually update to the
final protocol.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reusing the GPG signature check helpers we already have, verify
the signature in receive-pack and give the results to the hooks
via GIT_PUSH_CERT_{SIGNER,KEY,STATUS} environment variables.
Policy decisions, such as accepting or rejecting a good signature by
a key that is not fully trusted, is left to the hook and kept
outside of the core.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While signed tags and commits assert that the objects thusly signed
came from you, who signed these objects, there is not a good way to
assert that you wanted to have a particular object at the tip of a
particular branch. My signing v2.0.1 tag only means I want to call
the version v2.0.1, and it does not mean I want to push it out to my
'master' branch---it is likely that I only want it in 'maint', so
the signature on the object alone is insufficient.
The only assurance to you that 'maint' points at what I wanted to
place there comes from your trust on the hosting site and my
authentication with it, which cannot easily audited later.
Introduce a mechanism that allows you to sign a "push certificate"
(for the lack of better name) every time you push, asserting that
what object you are pushing to update which ref that used to point
at what other object. Think of it as a cryptographic protection for
ref updates, similar to signed tags/commits but working on an
orthogonal axis.
The basic flow based on this mechanism goes like this:
1. You push out your work with "git push --signed".
2. The sending side learns where the remote refs are as usual,
together with what protocol extension the receiving end
supports. If the receiving end does not advertise the protocol
extension "push-cert", an attempt to "git push --signed" fails.
Otherwise, a text file, that looks like the following, is
prepared in core:
certificate version 0.1
pusher Junio C Hamano <gitster@pobox.com> 1315427886 -0700
7339ca65... 21580ecb... refs/heads/master
3793ac56... 12850bec... refs/heads/next
The file begins with a few header lines, which may grow as we
gain more experience. The 'pusher' header records the name of
the signer (the value of user.signingkey configuration variable,
falling back to GIT_COMMITTER_{NAME|EMAIL}) and the time of the
certificate generation. After the header, a blank line follows,
followed by a copy of the protocol message lines.
Each line shows the old and the new object name at the tip of
the ref this push tries to update, in the way identical to how
the underlying "git push" protocol exchange tells the ref
updates to the receiving end (by recording the "old" object
name, the push certificate also protects against replaying). It
is expected that new command packet types other than the
old-new-refname kind will be included in push certificate in the
same way as would appear in the plain vanilla command packets in
unsigned pushes.
The user then is asked to sign this push certificate using GPG,
formatted in a way similar to how signed tag objects are signed,
and the result is sent to the other side (i.e. receive-pack).
In the protocol exchange, this step comes immediately before the
sender tells what the result of the push should be, which in
turn comes before it sends the pack data.
3. When the receiving end sees a push certificate, the certificate
is written out as a blob. The pre-receive hook can learn about
the certificate by checking GIT_PUSH_CERT environment variable,
which, if present, tells the object name of this blob, and make
the decision to allow or reject this push. Additionally, the
post-receive hook can also look at the certificate, which may be
a good place to log all the received certificates for later
audits.
Because a push certificate carry the same information as the usual
command packets in the protocol exchange, we can omit the latter
when a push certificate is in use and reduce the protocol overhead.
This however is not included in this patch to make it easier to
review (in other words, the series at this step should never be
released without the remainder of the series, as it implements an
interim protocol that will be incompatible with the final one).
As such, the documentation update for the protocol is left out of
this step.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Everywhere else we use PKT-LINE to denote the pkt-line formatted
data, but "shallow/deepen" messages are described with PKT_LINE().
Fix them.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our signed-tag objects set the standard format used by Git to store
GPG-signed payload (i.e. the payload followed by its detached
signature) [*1*], and it made sense to have a helper to find the
boundary between the payload and its signature in tag.c back then.
Newer code added later to parse other kinds of objects that learned
to use the same format to store GPG-signed payload (e.g. signed
commits), however, kept using the helper from the same location.
Move it to gpg-interface; the helper is no longer about signed tag,
but it is how our code and data interact with GPG.
[Reference]
*1* http://thread.gmane.org/gmane.linux.kernel/297998/focus=1383
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Earlier, ffb6d7d5 (Move commit GPG signature verification to
commit.c, 2013-03-31) moved this helper that used to be in pretty.c
(i.e. the output code path) to commit.c for better reusability.
It was a good first step in the right direction, but still suffers
from a myopic view that commits will be the only thing we would ever
want to sign---we would actually want to be able to reuse it even
wider.
The function interprets what GPG said; gpg-interface is obviously a
better place. Move it there.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We use it to make sure that the feature request is sent only once on
the very first request packet (ignoring the "shallow " line, which
was an unfortunate mistake we cannot retroactively fix with existing
receive-pack already deployed in the field) and we set it to "true"
with cmds_sent++, not because we care about the actual number of
updates sent but because it is merely an idiomatic way.
Set it explicitly to one to clarify that the code that uses this
variable only cares about its zero-ness.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The main loop over remote_refs list inspects the ref status
to see if we need to generate pack data (i.e. a delete-only push
does not need to send any additional data), resets it to "expecting
the status report" state, and formats the actual update commands
to be sent.
Split the former two out of the main loop, as it will become
conditional in later steps.
Besides, we should have code that does real thing here, before the
"Finally, tell the other end!" part ;-)
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The variable counts how many non-deleting command is being sent, but
is only checked with 0-ness to decide if we need to send the pack
data.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Similar to the previous one for send-pack, make it easier and
cleaner to add to capability advertisement.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A run of 'var ? " var" : ""' fed to a long printf string in a deeply
nested block was hard to read. Move it outside the loop and format
it into a strbuf.
As an added bonus, the trick to add "agent=<agent-name>" by using
two conditionals is replaced by a more readable version.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We tried to avoid sending one extra byte, NUL and nothing behind it
to signal there is no protocol capabilities being sent, on the first
command packet on the wire, but it just made the code look ugly.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A new helper function ref_update_to_be_sent() decides for each ref
if the update is to be sent based on the status previously set by
set_ref_status_for_push() and also if this is a mirrored push.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
20e8b465 (refactor ref status logic for pushing, 2010-01-08)
restructured the code to set status for each ref to be pushed, but
did not quite go far enough. We inspect the status set earlier by
set_refs_status_for_push() and then perform yet another update to
the status of a ref with an otherwise OK status to be deleted to
mark it with REF_STATUS_REJECT_NODELETE when the protocol tells us
never to delete.
Split the latter into a separate loop that comes before we enter the
per-ref loop. This way we would have one less condition to check in
the main loop.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make a helper function to accept a line of a protocol message and
queue an update command out of the code from read_head_info().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This piece of code reads object names of shallow boundaries, not
old_sha1[], i.e. the current value the ref points at, which is to be
replaced by what is in new_sha1[].
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Ideally, we should have also allowed the first "shallow" to carry
the feature request trailer, but that is water under the bridge
now. This makes the next step to factor out the queuing of commands
easier to review.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An "update" command in the protocol exchange consists of 40-hex old
object name, SP, 40-hex new object name, SP, and a refname, but the
first instance is further followed by a NUL with feature requests.
The command structure, which has a flex-array member that stores the
refname at the end, was allocated based on the whole length of the
update command, without excluding the trailing feature requests.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
CodingGuidelines states that the first #include in C files should be
git-compat-util.h or another header file that includes it, such as
cache.h or builtin.h.
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
glibc has deprecated the use of _BSD_SOURCE define
warning "_BSD_SOURCE and _SVID_SOURCE are deprecated, use _DEFAULT_SOURCE"
To make it easier to maintain a cross platform source code, that
warning can be suppressed by _DEFAULT_SOURCE.
Define both _BSD_SOURCE and _DEFAULT_SOURCE to clean-up the build.
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Call the functions behind git prune-packed and git update-server-info
directly instead of using run_command(). This is shorter, easier and
quicker.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We allocate pack information in a static global list but
never clean it up. This leaks memory, and means that calling
update_server_info twice will generate a buggy file (it will
have duplicate entries).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since "git update-server-info" may be called automatically
as part of a push or a "gc --auto", we should be robust
against two processes trying to update it simultaneously.
However, we currently use a fixed tempfile, which means that
two simultaneous writers may step on each other's toes and
end up renaming junk into place.
Let's instead switch to using a unique tempfile via mkstemp.
We do not want to use a lockfile here, because it's OK for
two writers to simultaneously update (one will "win" the
rename race, but that's OK; they should be writing the same
information).
While we're there, let's clean up a few other things:
1. Detect write errors. Report them and abort the update
if any are found.
2. Free path memory rather than leaking it (and clean up
the tempfile when necessary).
3. Use the pathdup functions consistently rather than
static buffers or manually calculated lengths.
This last one fixes a potential overflow of "infofile" in
update_info_packs (e.g., by putting large junk into
$GIT_OBJECT_DIRECTORY). However, this overflow was probably
not an interesting attack vector for two reasons:
a. The attacker would need to control the environment to
do this, in which case it was already game-over.
b. During its setup phase, git checks that the directory
actually exists, which means it is probably shorter
than PATH_MAX anyway.
Because both update_info_refs and update_info_packs share
these same failings (and largely duplicate each other), this
patch factors out the improved error-checking version into a
helper function.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We form all of our directories in a strbuf, but never release it.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On my Debian 7 system, this fixes annoying warnings when the output
of "git svn" commands are redirected:
Unable to get Terminal Size. The TIOCGWINSZ ioctl didn't work.
The COLUMNS and LINES environment variables didn't work. The
resize program didn't work.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Allow -B and -A to act as short aliases for --before and --after
options respectively. This reduces typing and hopefully allows
reuse of muscle memory for grep(1) users.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Git no longer seems to use these flags or their associated config keys;
when they are present, git-svn outputs a message indicating that they
are being ignored.
Signed-off-by: Lawrence Velázquez <vq@larryv.me>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Calling "git svn info $(pwd)" would hit:
"Reading from filehandle failed at ..."
errors due to improper prefixing and canonicalization.
Strip the toplevel path from absolute filesystem paths to ensure
downstream canonicalization routines are only exposed to paths
tracked in git (or SVN).
v2:
Thanks to Andrej Manduch for originally noticing the issue
and fixing my original version of this to handle
more corner cases such as "/path/to/top/../top" and
"/path/to/top/../top/file" as shown in the new test cases.
v3:
Fix pathname portability problems pointed out by Johannes Sixt
with a hint from brian m. carlson.
Cc: Johannes Sixt <j6t@kdbg.org>
Cc: "brian m. carlson" <sandals@crustytoothpaste.net>
Signed-off-by: Andrej Manduch <amanduch@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Commands such as "git svn init/fetch/dcommit" do not prompt for client
certificate/password if they are stored in SVN config file. Make
"git svn branch" consistent with the other commands, as SVN::Client is
capable of building its own authentication baton from information in the
SVN config directory.
Signed-off-by: Monard Vong <travelingsoul86@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
test_cmp is intended to produce diff output for human consumption. The
input in one instance in t9300-fast-import.sh are binary files, however.
Use test_cmp_bin to compare the files.
This was noticed because on Windows we have a special implementation of
test_cmp in pure bash code (to ignore differences due to intermittent CR
in actual output), and bash runs into an infinite loop due to the binary
nature of the input.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Our filesystem ref storage does not allow D/F conflicts; so
if "refs/heads/a/b" exists, we do not allow "refs/heads/a"
to exist (and vice versa). This falls out naturally for
loose refs, where the filesystem enforces the condition. But
for packed-refs, we have to make the check ourselves.
We do so by iterating over the entire packed-refs namespace
and checking whether each name creates a conflict. If you
have a very large number of refs, this is quite inefficient,
as you end up doing a large number of comparisons with
uninteresting bits of the ref tree (e.g., we know that all
of "refs/tags" is uninteresting in the example above, yet we
check each entry in it).
Instead, let's take advantage of the fact that we have the
packed refs stored as a trie of ref_entry structs. We can
find each component of the proposed refname as we walk
through the trie, checking for D/F conflicts as we go. For a
refname of depth N (i.e., 4 in the above example), we only
have to visit N nodes. And at each visit, we can binary
search the M names at that level, for a total complexity of
O(N lg M). ("M" is different at each level, of course, but
we can take the worst-case "M" as a bound).
In a pathological case of fetching 30,000 fresh refs into a
repository with 8.5 million refs, this dropped the time to
run "git fetch" from tens of minutes to ~30s.
This may also help smaller cases in which we check against
loose refs (which we do when renaming a ref), as we may
avoid a disk access for unrelated loose directories.
Note that the tests we add appear at first glance to be
redundant with what is already in t3210. However, the early
tests are not robust; they are run with reflogs turned on,
meaning that we are not actually testing
is_refname_available at all! The operations will still fail
because the reflogs will hit D/F conflicts in the
filesystem. To get a true test, we must turn off reflogs
(but we don't want to do so for the entire script, because
the point of turning them on was to cover some other cases).
Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With "hash-object --literally", write a tag object that is not
supposed to pass one of the new checks added to "fsck", and make
sure that the new check catches the breakage.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* js/fsck-tag-validation:
Make sure that index-pack --strict checks tag objects
Add regression tests for stricter tag fsck'ing
fsck: check tag objects' headers
Make sure fsck_commit_buffer() does not run out of the buffer
fsck_object(): allow passing object data separately from the object itself
Refactor type_from_string() to allow continuing after detecting an error
