Add constants for parsing git wire protocol version
This would allow other JGit users to access and reuse the constants.
Change-Id: I1608802f45586af5f8582afa592e26679e9cebe3
Signed-off-by: David Ostrovsky <david@ostrovsky.org>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Do not send empty blob in response to blob:none filter
If I create a repository containing an empty file and clone it
with
git clone --no-checkout --filter=blob:none \
https://url/of/repository
then I would expect no blobs to be transferred over the wire. Alas,
JGit rewrites filter=blob:none to filter=blob:limit=0, so if the
repository contains an empty file then the empty blob gets
transferred.
Fix it by teaching JGit about filters based on object type to
complement the existing filters based on object size. This prepares
us for other future filters such as object:none.
In particular, this means we do not need to look up the size of the
filtered blobs, which should speed up clones. Noticed by Anna
Pologova and Terry Parker.
Change-Id: Id4b234921a190c108d8be2c87f54dcbfa811602a
Signed-off-by: Jonathan Nieder <jrn@google.com>
Teach the FilterSpec serialization code about tree filters so they can
be communicated over the wire and understood by the server.
While we're here, harden the FilterSpec serialization code to throw
IllegalStateException if we encounter a FilterSpec that cannot be
expressed as a "filter" line. The only public API for creating a
Filterspec is to pass in a "filter" line to be parsed, so these should
not appear in practice.
Change-Id: I9664844059ffbc9c36eb829e2d860f198b9403a0
Signed-off-by: Jonathan Nieder <jrn@google.com>
UploadPack: Clear advertised ref map after negotiation
After negotiation phase of a fetch, the advertised ref map is no longer used and
can be safely cleared. For >1GiB repos object selection and packfile writing may
take 10s of minutes. For the chromium.googlesource.com/chromium/src repo, this
advertised ref map is >400MiB. Returning this memory to the Java heap is a major
scalability win.
Change-Id: I00d453c5ef47630c21f199e333e1cfcf47b7e92a
Signed-off-by: Minh Thai <mthai@google.com>
Since version 4.13 JUnit has an assertThrows method. Remove the
implementation in MoreAsserts and use the one from JUnit.
CQ: 21439
Change-Id: I086baa94aa3069cebe87c4cbf91ed1534523c6cb
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
The flag enabling sideband-all is used in two places: in UploadPack
for advertisement and in the protocol parser to read it from the
request.
This leds to problems in distributed deployments where the two requests of
a fetch can go to different servers with different configurations.
Use the existing allowsidebandall to accept the sideband-all request
(and respond to it) and introduce a new "advertisesidebandall" to toggle
the advertising of the feature.
Change-Id: I892d541bc3f321606c89bad1d333b079dce6b5fa
Signed-off-by: Ivan Frade <ifrade@google.com>
CachedPackUriProvider: Add size to the pack information
The object identifying packfiles to send them via packfile-uri contains
only the uri and the hash. This is the information that goes through the
wire. It would be useful to know also the size of those packfile, for
example to track how many bytes have been offloaded to HTTP.
Add size field the CachedPackUriProvider.PackInfo object.
Change-Id: If6b921b48a4764d936141c777879b148cc80bbd3
Signed-off-by: Ivan Frade <ifrade@google.com>
UploadPackTest: Stop using deprecated PacketLineIn constants
The DELIM and END constants are deprecated and using them causes
warnings. Replace them with the accessor methods.
Change-Id: Iadb27000755e8fd8c61d9218591f9d110b8265c8
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
UploadPackTest: Move "reachability" tests to its own file
UploadPackTest is already too long and it is covering too many aspects
of UploadPack. This makes difficult to see what is tests and if all
cases are covered.
Move the reachability-related tests to its own file. This moves also an
auxiliary function, reducing the length of UploadPack. Complete also the
coverage, adding combinations of bitmap availability/commits or
blobs/reachable or not.
Change-Id: Id5cfc9d0118d997da30e3886c91db996a86250fc
Signed-off-by: Ivan Frade <ifrade@google.com>
UploadPack: support custom packfile-to-URI mapping
Teach UploadPack to take a provider of URIs corresponding to cached
packs. When fetching, if the client supports the packfile-uri feature,
and if such a cached pack were to be streamed, instead send the
corresponding URI.
This packfile-uri feature is implemented in the jt/fetch-cdn-offload
branch of Git. There is interest in this feature [1], but it is not yet
merged.
[1] https://public-inbox.org/git/cover.1552073690.git.jonathantanmy@google.com/
Change-Id: I9a32dae131c9c56ad2ff4a8a9638ae3b5e44dc15
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
In a subsequent patch, in some cases, PackWriter#writePack will be
responsible for both the "packfile-uris" and "packfile" sections,
meaning that (in these cases) it must write the "packfile" section
header itself.
In preparation for that patch, move the writing of the "packfile"
section header closer to the invocation of PackWriter#writePack when the
entire fetch response is configured to use the sideband. This means that
"packfile" is written *after* objects are counted (and progress messages
sent to the client in sideband 2) when the "sideband-all" feature is
used (whether "packfile-uris" is used or not), and written *before*
objects are counted otherwise.
Having code to write "packfile" in two places is unfortunate but
necessary. When "sideband-all" is not used, object counting has to
happen after "packfile" is written, because "packfile" activates the
sideband that allows counting progress to be transmitted. When
"packfile-uris" is used, object counting has to happen before "packfile"
is written, because object counting determines whether to send
"packfile-uris" or "packfile". When "sideband-all" is used but
"packfile-uris" is not used, either way works; this commit uses
"packfile-uris" behavior in this case.
Also make the naming of the sideband-activating methods in PacketLineOut
more consistent.
Change-Id: Ifbfd26cc26af10c41b77758168833702d6983df1
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Use a Consumer instead of several nullable variables to further
configure UploadPack. This is in preparation for a test in a subsequent
patch needing further customization of the UploadPack object before
invoking it.
Change-Id: I074dff92c711a5ba74558bb4b06c42c115fb9b7f
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Allow the client to specify "sideband-all" in a fetch v2 request,
indicating that the whole response is to be multiplexed (with a sideband
indicator on every non-flush and non-delim pkt) instead of only the
packfile being multiplexed. This allows, for example, progress messages
to be sent at any point in the response.
This implements the "sideband-all" feature documented in
Documentation/technical/protocol-v2.txt in Git.
Change-Id: I3e7f21c88ff0982b1b7ebb09c9ad6c742c4483c8
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
UploadPackTest.java contains tests that check behavior when
"allowfilter" and "allowrefinwant" are not set, are set, and are not set
but the client insists on using them anyway. Because another capability
is to be included in a subsequent patch, refactor the common code in
these tests.
Remove setBoolean calls with "false", as they are no-ops.
Also take the opportunity to eliminate the overspecification of the
"fetch=" line returned by the capability advertisement.
Change-Id: I289bbd11c902a513cd8d53bc34767e61ebbd5f17
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
By doing this, exceptions thrown by sendPack are also covered by the
same code.
Change-Id: I3509f2d832af1410f307e931577e4d07e32b014e
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
This is a new assertion that will be introduced in JUnit 4.13. Unlike
ExpectedException rule, this makes it easy to test other aspects of the
thrown exception, such like ServiceMayNotContinueException's status
code. Introduce this as before making changes to UploadPackTest more.
Change-Id: Ied7b3071ffcd0e93eece35b01e0abc5ff65645f2
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
If a client clones with "--filter=blob:none", the checkout that "git
clone" automatically does causes the client to fetch all blobs at HEAD.
When fetching from a non-bitmapped repository, this will fail if an
object walk is ever needed, because JGit currently rejects such requests
- see the commit message of d3021788d2 ("Use bitmaps for non-commit
reachability checks", 2017-11-10) for more information.
Rejecting such requests in the absence of bitmaps is probably
overzealous: it is true that the server would prefer to have bitmaps in
this case, but there might be a small proportion of repos (for example,
very small repos or newly created ones) that do not have bitmaps, yet
the server would still like to have partial clones for them.
So, allow such requests, performing the object walk reachability check
if necessary. Limit this to servers with "uploadpack.allowFilter"
configured, so that servers wanting to support partial clone have this
functionality, and servers that do not support partial clone do not have
to pay the object walk reachability check cost.
Change-Id: I51964bafec68696a799625d627615b4f45ddbbbf
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
When cloning repository with --single-branch option, tag chains are not
packed and pack file is broken in some cases.
Typical test-case:
git tag -a test_tag <commit-id>
git tag -a test_prev_tag test_tag
git tag -d test_tag
git clone --single-branch <repository>
fatal: did not receive expected object <test_tag_id>
The reason for that is missing object for original test_tag reference,
which was deleted.
Problem description:
When pack-objects is given --include-tag, it peels each tag reference
down to a commit. If the commit is prepared to be packed, we we have to
include such tag too. The problem is when the tag points to through some
chain of other tag to commit. Then, the inner tags are not added leading
to broken pack.
Fix:
When going to commit, we have to check and add any of the tags on the
way (if they were not selected, which may happen with --single-branch
option).
Change-Id: I1682d4a2c52d674f90a1b021e0f6c3524c5ce5bc
Signed-off-by: Pavel Flaška <Pavel.Flaska@gmail.com>
Signed-off-by: Matthias Sohn <matthias.sohn@sap.com>
Replace most usages of PacketLineIn.END with PacketLineIn.end()
PacketLineIn.END is only referenced in tests. Replace most of those
with a new package visible end() method.
Remaining usages of PacketLineIn.END are in the form:
while ((line = pckIn.readString()) != PacketLineIn.END) {
and are not trivial replacements, hence are not touched in this change.
Change-Id: Id77c5321ddcad127130b246bde8f08736e60e1ea
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Deprecate DELIM with the intention of making it private in a future
release.
Callers that want to test if a packet line string is the delimiter
should use the isDelimiter(String) method.
The only other references to DELIM in the JGit code are in tests. For
those, introduce a package visible delimiter() method.
Change-Id: I21e8bbac0ffb9ef710c9753e23435416b09a4891
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Replace trivial reference comparison of PacketLineIn.{DELIM,END}
Replace reference comparisons of PacketLineIn's DELIM and END strings
with usage of the helper methods isDelimiter() and isEnd().
Change-Id: I52dcfc4ee9097f1bd6970601c716701847d9eebd
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
If a tree is visited during pack and filtered out with tree:<depth>, we
may need to include it if it is visited again at a lower depth.
Until now we revisit it no matter what the depth is. Now, avoid
visiting it if it has been visited at a lower or equal depth.
Change-Id: I68cc1d08f1999a8336684a05fe16e7ae51898866
Signed-off-by: Matthew DeVore <matvore@gmail.com>
tree:<depth> should not traverse overly-deep trees
If we are traversing a tree which is too deep, then there is no need to
traverse the children. Skipping children is much faster than traversing
the possibly thousands of objects which are directly or indirectly
referenced by the tree.
Change-Id: I6d68cc1d35da48e3288b9cc80356a281ab36863d
Signed-off-by: Matthew DeVore <matvore@gmail.com>
This is used when fetching, and in particular to populate a partial
clone or a virtual file system cache as the user navigates. With this,
a client can pre-fetch a few directories deeper than only the current
directory.
depth:0 will omit all trees, and is useful if you only want to fetch
the commits of a repository, or fetch just a single tree or blob object.
depth:1 will fetch only the root tree of all commits fetched. depth:2
will fetch the root tree and all blobs and tree objects directly
referenced from it. depth:3 gets one more level, and so on. depth:#
will not filter a blob or tree that is explicitly marked wanted.
Bitmaps are disabled when this filter is used.
This implementation is quite slow because it iterates over all omitted
objects rather than skipping them. This will be addressed in follow-up
commits.
Change-Id: Ic312fee22d60e32cfcad59da56980e90ae2cae6a
Signed-off-by: Matthew DeVore <matvore@gmail.com>
UploadPackTest: Stop using deprecated Transport.setFilterBlobLimit(long)
Replace usage with the recommended setFilterSpec(FilterSpec).
Change-Id: Icc528d175f25234eeb2daa6b4c29a67a7a6d1e0a
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
Its implementation contains
} catch (IOException e) {
// Legacy API, assume error means "no"
return false;
}
Better to use ObjectDatabase#has, which throws IOException to report
errors.
Change-Id: I7de02f7ceb8f57b2a8ebdb16d2aa4376775ff933
Signed-off-by: Jonathan Nieder <jrn@google.com>
Suppose that a repository has the following commit graph:
B C
\ /
A
and it was cloned with --shallow-exclude=A. DepthGenerator does not mark
C as shallow, causing an invalid repository to be produced on the
client, because A is not sent. (A similar issue occurs when
--shallow-since is used to exclude A but neither B nor C.)
This happens whenever an excluded commit has more than one child that is
to be sent to the client. Fix DepthGenerator to handle this case
correctly.
While we're editing DepthWalk.Commit, fix the documentation of
DepthWalk.Commit#isBoundary.
Change-Id: I7068abf0fe0c864d1b0e56e1616dad1aa8719411
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
This allows clients to use the --shallow-exclude parameter (producing a
"deepen-not <ref>" line when communicating with the server) in their fetch
commands when fetching against a JGit server using protocol v2.
Note that the implementation in this commit is somewhat inefficient, as
described in the TODO comment in DepthGenerator.
Change-Id: I9fad3ed9276b624d8f668356ffd99a067dc67ef7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Throw error when deepen-since excludes all commits
In C Git, when a client fetches with "git fetch --shallow-since=<date>
origin <ref>", and all commits reachable from <ref> are older than
<date>, the server dies with a message "no commits selected for shallow
requests". That is, (1) the --shallow-since filter applies to the commit
pointed to by the ref itself, and (2) there is a check that at least one
commit is not filtered out. (The pack-protocol.txt documentation does
not describe this, but the C implementation does this.)
The implementation in commit 1bb430dc21 ("UploadPack: support
deepen-since in protocol v2", 2018-09-27) does neither (1) nor (2), so
do both of these.
Change-Id: I9946327a71627626ecce34ca2d017d2add8867fc
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
UploadPack: Return correct peer user agent on v2 requests
UploadPack.getPeerUserAgent() doesn't produce the expected results for
protocol v2 requests. In v2, the agent reported in the request (in an
"agent=" line) is not in the clientCapabilities but in a field on its
own. This makes getPeerUserAgent default to the transport user agent.
Making "agent" a shared property between protocol v0/v1 and v2 fixes the
problem, simplifies the function and harmonizes the implementation
between protocol versions.
In a follow up commit the "agent" will be identified on parsing time,
instead of taking it from the client capabilities.
Change-Id: Idf9825ec4e0b81a1458c8e3701f3e28aafd8a32a
Signed-off-by: Ivan Frade <ifrade@google.com>
Accept protocol v2 server options on fetch and ls-refs requests
In protocol v2, a command request can be followed by server options
(lines like "agent=<>" and "server-option=<>"), but current code
doesn't accept those lines.
Advertise the "server-option" capability, parse the lines and add
them to the request objects.
Other code in JGit can see this options and act accordingly via the
protocol v2 hooks.
This should not require any change in the client side.
Change-Id: If3946390f9cc02d29644b6ca52534b6f757bda9f
Signed-off-by: Ivan Frade <ifrade@google.com>
Support the deepen-since parameter when requested by a client using
protocol v2. This is done by:
- adding a DepthWalk.RevWalk#setDeepenSince method
- updating DepthGenerator to recognize when deepen-since is set
- recording in DepthWalk.Commit whether a commit is a boundary commit
Existing users of DepthWalk such as UploadPack previously recognized
boundary commits by comparing their depths against the threshold, not
tracking whether any parents were truly excluded. This behavior is
preserved - UploadPack considers a commit as boundary if its depth is
equal to the threshold *or* a parent was excluded (whether by depth or
by deepen-since).
Change-Id: I852bba6b1279f9cc8aee38282e9339d62b8dcddc
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
UploadPackTest: Avoid unnecessarily boxing int into Integer
The statement:
assertThat(recvStream.available(), is(0));
results in a warning from Eclipse:
The expression of type int is boxed into Integer
because recvStream.available() returns int, but the hamcrest is()
method takes an Integer.
Replace it with the equivalent JUnit assertion.
Also remove the suppression of another similar warning and fix that
in the same way.
Change-Id: I6f18b304a540bcd0a10aec7d3abc7dc6f047fe80
Signed-off-by: David Pursehouse <david.pursehouse@gmail.com>
UploadPack: avoid conflating shallow commit lists in protocol v2
At the moment there are two copies of the client shallow commit list:
one in the request and another in the clientShallowCommits member of
the class.
The verifyShallowCommit function was removing missing object ids
from the member but not the request list, and code afterwards was
using the request's version.
In practice, this didn't cause trouble because these shallow commits
are used as endpoint for a walk, and missing ids are just never reached.
Change-Id: I70a8f1fd46de135da09f16e5d954693c8438ffcb
Signed-off-by: Ivan Frade <ifrade@google.com>
In Git protocol v2, UploadPack and ReceivePack have the same
capabilities and can process any protocol v2 request. For example, a
client can sent a "fetch" command to the "/git-receive-pack" endpoint.
This makes it difficult for existing hook interfaces. For example,
PreUploadHook takes UploadPack, but a "fetch" command may be received by
ReceivePack.
To resolve this skew, this change introduce a different hook interface
for the protocol v2. The hook takes a request that is independent to the
handlers (UploadPack, ReceivePack). Also this makes it clear what
parameters the hook is counting on, instead of keep track of the hook
using getters from UploadPack / ReceivePack.
Bug: 534847
Change-Id: I71f3266584483db1e2b2edfc1a72d0bdf1bb6041
Signed-off-by: Masaya Suzuki <masayasuzuki@google.com>
When fetching with protocol v2, git expects the shallow-info section to
appear before wanted-refs if both appear in the response. Teach
UploadPack to do this.
Change-Id: Ie26a91edcce5d27a1d727d7fba5c30e1144e118b
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
UploadPack already allows the client to send wanted OIDs as "want"
lines. Extend UploadPack to also allow the client to send wanted ref
names as "want-ref" lines when the fetch is done using protocol v2.
The corresponding Git commit is 516e2b76bd ("upload-pack: implement
ref-in-want", 2018-06-28).
To support a two-stage rollout, two configuration variables are
provided: uploadpack.allowrefinwant (default "false") allows clients to
specify "want-ref" in their requests, and uploadpack.advertiserefinwant
(default "true") makes UploadPack advertise this capability. If
uploadpack.allowrefinwant is true but uploadpack.advertiserefinwant is
false, UploadPack will not advertise that it supports "want-ref", but it
will support it.
Change-Id: I3c24077949640d453af90d81a7f48ce4b8ac9833
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
When processing a fetch using protocol v2, UploadPack#fetchV2 sends an
extraneous flush pkt when also sending a packfile (#sendPack sending its
own flush pkt). Update that method to only send the flush pkt if the
packfile is not being sent.
Change-Id: I7117a264bccd2d7f3a048645fcb8425a9d78d526
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Always send refs' objects despite "filter" in pack
In a0c9016abd ("upload-pack: send refs' objects despite "filter"",
2018-07-09), Git updated the "filter" option in the fetch-pack
upload-pack protocol to not filter objects explicitly specified in
"want" lines, even if they match the criterion of the filter. Update
JGit to match that behavior.
Change-Id: Ia4d74326edb89e61062e397e05483298c50f9232
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
If the configuration variable uploadpack.allowfilter is true, advertise
that "filter" is supported, and support it if the client sends such an
argument.
Change-Id: I7de66c0a0ada46ff71c5ba124d4ffa7c47254c3b
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
A subsequent patch will dynamically generate the capability
advertisement, so the capability advertisements produced are not always
the same. Separate the checking of the advertisements into its own test
method.
Change-Id: I768d14b9d1a244d5d886c42ffd62ef3957b518fb
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
JGit's implementation of the fetch command of protocol v2, unlike its
implementation of ls-refs, currently tolerates unknown arguments.
Tighten fetch to not allow unrecognized arguments and add tests to
verify this behavior for both ls-refs and fetch.
Change-Id: I321161d568bd638252fab1a47b06b924d472a669
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Add support for the "shallow" and "deepen" parameters in the "fetch"
command in the fetch-pack/upload-pack protocol v2. Advertise support for
this in the capability advertisement.
TODO: implement deepen-relative, deepen-since, deepen-not
Change-Id: I7ffd80d6c38872f9d713ac7d6e0412106b3766d7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Jonathan Nieder <jrn@google.com>
Add support for the "ofs-delta" parameter in the "fetch" command in
the fetch-pack/upload-pack protocol v2.
Change-Id: I728cf986082fce4ddeb6a6435897692e15e60cc7
Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
David Ostrovsky
Jonathan Nieder
Minh Thai
Matthias Sohn
David Pursehouse
Ivan Frade
Jonathan Tan
Masaya Suzuki
Pavel Flaška
Carsten Hammer
Matthew DeVore