Summary:
We want to enable pushing tree packs during pushrebase, and eventually we can
stop including flat manifest information entirely. The first step is to generate
the tree packs during push. A future patch will handle the incoming parts.
Test Plan: Ran the tests. A future patch adds pack file handling to test this
Reviewers: #mercurial, mjpieters
Reviewed By: mjpieters
Subscribers: mitrandir, medson, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D5329775
Signature: t1:5329775:1499705168:43ea77c9a5c85c09b40910e102535d89a0504a4d
Summary:
Improve the error message from the blocknonpushrebase hook to inform the user
that they need the pushrebase extension enabled.
Previously the message indicated that you mush push using
`hg push --to <bookmark>`, but this is not sufficient. Enabling the pushrebase
hook is the most important step here. (The `--to` flag is not part of vanilla
mercurial, and is added by pushrebase, but other extensions like remotenames
also add a `--to` flag.)
Test Plan:
Added a unit tests for `hg push --to <bookmark>` with remotenames enabled but
pushrebase disabled.
Reviewers: #mercurial, davidsp, zhihuih
Reviewed By: zhihuih
Subscribers: quark, medson, mjpieters, net-systems-diffs@fb.com
Differential Revision: https://phabricator.intern.facebook.com/D5276550
Signature: t1:5276550:1497899421:3656bcc3e7358f97fc852e14307b34d8aa4d9951
Before this patch, the pretxnchangegroup 'node' argument would be the last
added node whereas mercurial documention specifies that it should be the first
new changeset. The 'node_last' argument has also been added to both
pretxnchangegroup and changegroup hooks as per mercurial documentation.
This helps doing some validation before accepting a pushrebase.
For example, if you want to prevent users to accidentally pushrebase
changesets to the wrong branch by checking if all pushed changesets are
on the same branch as the destination (--to)
Summary: Upstream has deprecated cmdutil.commands() in favor of registrar.commands()
Test Plan: Ran the tests
Reviewers: #mercurial, quark
Reviewed By: quark
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D5106486
Signature: t1:5106486:1495485074:0e20f00622cc651e8c9dda837f84dd84cc51099e
Summary:
D4865150 and D4934720 aren't effective in our current setup. The direct
cause in the code is because the server couldn't find common marker version:
```
# old server-side code, returns empty in our current setup
obsolete.commonversion(bundle2.obsmarkersversion(reply.capabilities))
```
Upon investigation, it's because there is no `exchange` enabled client-side.
But we do want one-way (server->client) markers for the rebased commits, as
long as obsstore is enabled (createmarkers is set, without exchange).
The upstream expects the server to have obsstore enabled, and exchange
enabled, to send markers. Since we are generating markers without an
obsstore (see D4865150), we are on our own way. This diff makes it one step
further.
This diff adds an explicit parameter to the `b2x:rebase` part to tell the
server what obsmarker format the client supports so the server could make a
right decision without relying on the "standard" `reply.capabilities`, which
is affected by the exchange option.
Test Plan: Change the existing test, make sure the old code fails.
Reviewers: #mercurial, durham
Reviewed By: durham
Subscribers: durham, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4997972
Signature: t1:4997972:1493848751:14c29654b2e8246bd12a8de8820af5b3773e2fb7
Summary:
stateful chg will add a new parameter to repo initialization. This diff makes
sure pushrebase works with that change.
Test Plan: codereview
Reviewers: #mercurial, durham
Reviewed By: durham
Subscribers: durham, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4978687
Signature: t1:4978687:1493689190:7119205d76a8ac26dbf2bcf0be9c74ced9fc16b4
Summary:
Previously we ignore "draft -> public" phase move, and new nodes returned by
the server may remain "draft" until the next pull. This diff records the
obsmarkers and updates phases of the new nodes correctly so they become
public without an extra hg pull.
Test Plan: Modified an existing testcase
Reviewers: #mercurial, durham, rmcelroy
Reviewed By: durham
Subscribers: durham, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4934720
Signature: t1:4934720:1493076241:9b23da1140b68aece9e838c604e64611b3498794
Summary:
Previously when the client has obsstore enabled, and the server has obsstore
disabled, clients won't get the obsmarkers, which is suboptimal. This diff
makes the server send obsmarkers in that case.
Practically, this means people will no longer need to run `strip` after landing
a diff in this repo.
Note: `test-pushrebase-manifests.t` is somehow flaky about the empty lines. I
guess they are not related to the change. But the test change is not 100%
reproducible.
Test Plan: Added a new test case
Reviewers: #mercurial, stash
Reviewed By: stash
Subscribers: rmcelroy, ikostia, stash, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4865150
Signature: t1:4865150:1491895844:afbc3079a40a9a9fa9af1eab4eeaca91091e8d2d
Summary:
Previously we copied the old bundle dirmancache entries into the new bundle
dirmancache entries, but that missed any manifests that weren't already in the
cache. Since we have a list of bundle revs that are guaranteed to be needed
during the rebase, lets read them before the lock and force them into the cache
after the lock.
Test Plan:
Ran all pushrebase tests with a block like:
```
bundle.manifestlog[mfnode].read()
realmf = newdirmancache[''][mfnode].read()
if realmf.text() != mfdict.text():
raise Exception("cache invalid")
```
...to verify the manifest we cached is the same as can be read from disk.
Reviewers: #mercurial, durham
Reviewed By: durham
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4774728
Signature: t1:4774728:1490736995:6e20d69181b4df76aeb800ea246a0dae0a4b802d
Summary:
This fetches the old master's full text before the lock is taken, and puts it in
the revlog._cache field so we don't have to read the full delta chain when we go
to read the latest master revision. This minimizes the amount of time we have to
spend processing the manifest during the first write.
Test Plan:
Ran all tests using a block like the following:
```
for mfnode, mfdict in bundlerepocache.iteritems():
bundle.manifestlog[mfnode].read()
existing = newdirmancache[""][mfnode].read().text()
new = mfdict.text()
if new != existing:
raise Exception("Cache failed")
```
...which raises if the data we were about to pull from the cache was different from what we were reading from the manifest before.
Reviewers: #mercurial, durham, rmcelroy
Reviewed By: durham
Subscribers: rmcelroy, quark, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4774716
Signature: t1:4774716:1490735677:85bac79999d607f5554d090978d0ee6787f5e11c
Summary:
Previously, pushrebase detected if it was ok to rebase a commit by diff'ing the
original common ancestor of the bundle and master with the new master. This
required reading both manifests. To improve performance, let's read the
changelog files list instead, which is much, much faster.
Merge commits don't have a files list that is representative of the actual diff,
so fall back to manifest diff in that case.
Test Plan: ran tests
Reviewers: #mercurial, stash, durham
Reviewed By: stash, durham
Subscribers: stash, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4774694
Signature: t1:4774694:1490601056:18f78ab7c3e35a1beb2a7775160271a3311e38bb
Summary: `resource` is UNIX-only module, we can't use it if we want `pushrebase` to work on Windows. As we only use `getpagesize()`, we can also do `mmap.PAGESIZE` instead.
Test Plan: - apply change on Linux, see that they do not introduce any new failures
Reviewers: #sourcecontrol, durham, rmcelroy
Reviewed By: rmcelroy
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4762553
Signature: t1:4762553:1490294453:fdcb2e9caa38f5d14eafc6ff36a1bb7d9429b6e9
Summary:
Upstream has added a new match argument to manifest.diff() and removed the
existing manifest.matches() function, so we need to update our internal usage.
A separate diff will update treemanifest to support the new diff() api.
Test Plan:
Ran the tests, some still fail because of the upstream changes, but
future patches fix those.
Reviewers: #mercurial
Differential Revision: https://phabricator.intern.facebook.com/D4677002
Summary: bundle file should be closed to ensure that all the temporary files were deleted
Test Plan: arc unit
Reviewers: #sourcecontrol, ikostia
Reviewed By: ikostia
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4377148
Signature: t1:4377148:1483451990:2a1362d9e524a54ffedad8f2bc09fc622efe8ad4
Summary:
test-check-code.t was updated in core mercurial so we have to fix it in
fb-hgext
Test Plan: arc unit
Reviewers: #sourcecontrol, quark
Reviewed By: quark
Subscribers: quark, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4377208
Signature: t1:4377208:1483451688:bc382cdab07c6d6127d67c902458a2ccfa1373d1
Summary:
Previously, if during a pushrebaes a merge commit was rebased and the
parent commit that contained the original rebase destinations content (i.e.
master's content) was on the p2 side, it would all be thrown away.
This is because the pushrebase commit works by copying the destination-repo-p1, then
applying changes from the diff of the original-repo-p1 vs original-repo-commit.
Since all the rebase destination contents was in destination-repo-p2, it is all
lost here.
The solution is to track which commit is the latest one to contain the
destination bookmark contents, and make sure it is always p1 when rebasing merge
commits. This will ensure that the destination bookmark contents are preserved
all the way up to the new commit for that bookmark.
Test Plan:
Added a test. It fails before the change and passes after. Another
test was also affected, since the p1/p2 ordering is different.
Reviewers: #mercurial, simonfar
Reviewed By: simonfar
Subscribers: simonfar, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4336427
Tasks: 15092414
Signature: t1:4336427:1481881555:a22b8c363f2f7a197cef192abf34bfca926ecdc2
Summary:
Using `testedwith = 'internal'` is not a good habit [1]. Having it
auto-updated in batch would also introduce a lot of churn. This diff makes
them "ships-with-fb-hgext". If we do want to fill the ideal "testedwith"
information, we could put it in a centric place, like a "fbtestedwith"
extension rewriting those "ships-with-fb-hgext" on the fly.
Maybe having in-repo tags for tested Mercurial releases is also a good idea.
[1]: www.mercurial-scm.org/repo/hg/rev/2af1014c2534
Test Plan: `arc lint`
Reviewers: #sourcecontrol, rmcelroy
Reviewed By: rmcelroy
Subscribers: rmcelroy, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4244689
Signature: t1:4244689:1480440027:3dc18d017b48beba1176fbfd120351889259eb4b
Summary:
Upstream has refactored the manifest class, so we need to update pushrebase to
work with the new structure. In particular, pushrebase takes advantage of the
manifest cache, so we need to carefully reconstruct the appropriate cache on the
bundle repo. We can't just .copy() it now because the cache contains manifestctx
instead of manifestdict, and the manifestctx contains a reference to the source
repo.
Test Plan: Ran the tests
Reviewers: #mercurial, quark
Reviewed By: quark
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D4180885
Signature: t1:4180885:1479228506:a0ff114e158caab25da22f6116e5a4090f5515c0
Summary:
"changeset" is a more official term and let's use it. Note that this patch
only changes documentation / i18n messages visible to the users and header
comment blocks to developers. Other places like comments in the code are
untouched.
With the "dialect" extension enabled, users will still see the more friendly
term - "commit".
Test Plan:
`arc unit`. Note the remotefilelog failure is probably unrelated - seems
related to ongoing / upcoming manifest refactoring upstream.
Reviewers: #sourcecontrol, rmcelroy
Reviewed By: rmcelroy
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D3900394
Signature: t1:3900394:1474470348:6a1b5691e2599cc47df18b227d56d1f9d3c7c906
Summary:
In upstream mercurial, 3d5e0fd41e0e changed discovery.outgoing() from taking a
changelog to taking the full repo object.
Test Plan: run tests, fewer stacktraces
Reviewers: #mercurial, ttung, durham, quark
Reviewed By: quark
Subscribers: mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D3819300
Signature: t1:3819300:1473090807:fd0e4c2956fee4ad7d45051b6b7c5be481c766f1
Summary:
Be a better citizen under system python path.
Fix all tests issues and change setup.py to use glob pattern to include
all extensions.
Test Plan:
Run tests and `make local`.
Also build and install the package and run `hg sl` in major repos.
Reviewers: #mercurial, ttung, rmcelroy
Reviewed By: rmcelroy
Subscribers: rmcelroy, durham, mjpieters
Differential Revision: https://phabricator.intern.facebook.com/D3534311
Signature: t1:3534311:1468275426:fe122646c8bd6c541e1889e73e9df28f86747ff2
