Summary:
Since reponame is part of the primary key, it also makes sense to add it to
WHERE clause. Otherwise query may trigger full table scan and it can be very
slow.
Test Plan: Run unittest
Reviewers: #fbhgext, durham
Reviewed By: #fbhgext, durham
Differential Revision: https://phab.mercurial-scm.org/D1256
`.pyc` files could cause false positive. Scan `.py` and `.pyx` files
explicitly.
Test Plan:
Make sure `uncommit.pyc` will not break check-code but `uncommit.pyx`,
`uncommit.py` will.
Differential Revision: https://phab.mercurial-scm.org/D1283
With the multidest rebase (D470), restack could be implemented as specifying
the source (`orphan() - obsolete()`) and a revset specifying destination for
each source revision.
This patch changes restack implementation to use that. The revset is
implemented as a private function `_destrestack`, like `_destrebase` in
`rebase.py`.
Most test changes are because the topo-sort in `rebase.py` removes some
unnecessary steps. So the resulting revision numbers are smaller.
There is one interesting test case that gets changed:
```
D
| # amend: B1 -> B2 -> B1
B2 B1 B3 # amend: B1 -> B3
\ | /
A
```
Previously, restack will move D to be on top of B3. Now restack will not
move it because `D` is considered stable (not orphaned) since none of its
ancestors are obsoleted. The new behavior seems to be more desirable.
More tests are added to test restack source revisions (should be in a same
stack) and some "prune" cases.
The transaction logic is also made more robust so `test-copytrace-amend.t`
does not cause a broken rebase state.
Since the rebase operation runs in a same rebase state, restack can be
continued via `rebase --continue` correctly after resolving merge conflicts.
This is reflected in a newly added test.
Test Plan:
With hg-dev, run `./script/unit.py`
Differential Revision: https://phab.mercurial-scm.org/D1139
test-remotefilelog-clone.t is flaky, and fails sporadically as follows:
$ hg clone --noupdate ssh://user@dummy/shallow full
streaming all changes
+ abort: unexpected response from remote server: empty string
remote: abort: Cannot clone from a shallow repo to a full repo.
- abort: unexpected response from remote server: empty string
[255]
The issue happens because the line starting "abort" and the line starting
"remote: abort" come from different processes, and there's no way to
guarantee the order in which they are output.
This change attempts to address the issue by capturing the output of the
local process and examining it separately.
Test Plan:
rt test-remotefilelog-clone.t
Differential Revision: https://phab.mercurial-scm.org/D1279
If we were using fastmanifest in a treeonly repository, hg show would try to
access the manifest delta via the revlog which would crash. The fix is to make
fastmanifest first try accessing the data via the tree, before falling back to
the delta.
My attempts at automatically testing this failed, but I had a repro in a real
repository and confirmed this fixed the issue.
There was a flakey test because the remote stderr output was getting eaten part
of the time. This fixes it by capturing all the output and then sending it to
stderr manually.
Tested it by running all the tests in parallel multiple times and not seeing the
issue anymore.
Also we now send that output to the debug output. We're already throwing a more
useful error message, so let's hide the noise.
Differential Revision: https://phab.mercurial-scm.org/D1276
Adds a test and improves output for dealing pulling a normal scratch branch into
a treeonly repo (so the manifest doesn't exist).
In the future we could fix this by having the server generate the trees on the
fly for infinitepush branches that don't already have them.
Differential Revision: https://phab.mercurial-scm.org/D1268
This makes the client side pushes also include tree packs in the push.
Also adds a test to see that the tree parts round trip between two tree-only
clients.
This was reverted in D921 because clients couldn't push old non-tree commits to
infinitepush anymore. I've fixed the issue and added a test to this commit.
Differential Revision: https://phab.mercurial-scm.org/D1017
The old version of this test relied on the sendflat config being set to false so
it would only send the trees. Since that config is gone, the test then started
sending flat manifests, so it wasn't actually testing what it should've been.
This patch fixes that by making it push a tree only commit and ensuring the
commit hook executes correctly. The test changed a bit, because the hash changed
after the push, because it was a tree only commit on the client (and therefore
used the tree hash) but a hybrid commit on the server (and therefore used the
flat hash).
Differential Revision: https://phab.mercurial-scm.org/D1267
Previously the sendflat config decided if we should send flat manifests or not.
To reduce the number of config permutations we have to handle, let's change the
code to always send the flat manifests, if it exists.
Differential Revision: https://phab.mercurial-scm.org/D1266
Previously we were relying on config values to determine when to send or not
send tree and flat manifests. This resulted in complicated permutations to test.
Let's remove these config values and always send trees if we have them. A future
patch will do the same for flat manifests.
Differential Revision: https://phab.mercurial-scm.org/D1265
This data might be useful too.
Move to _getavailablepackfiles so refresh() picks it up too.
Differential Revision: https://phab.mercurial-scm.org/D1281
There are some places that both fbsparse and sparse are enabled. To help
them migrate smoothly, let's disable sparse automatically if fbsparse is
detected.
Test Plan:
Run:
hg --config extensions.fbsparse --config extensions.sparse= sparse -h
Make sure it does not complain about:
extension 'sparse' overrides commands: ^sparse
Differential Revision: https://phab.mercurial-scm.org/D1273
Previously, hiddenoverride only unpins nodes when createmarkers gets called.
However, with scmutils.cleanupnodes, createmarkers will not get called when
the nodes are already obsoleted. This patch makes hiddenoverride check
cleanupnodes too to unpin nodes properly.
Differential Revision: https://phab.mercurial-scm.org/D994
This gives us more context about what pack file is problematic. It's useful
to detect issues like mmapping a 0-sized file.
Test Plan:
Make sure the code compiles without warnings. Run related tests.
```
make local
rt test-cstore.t
```
Make sure `setup.py` specifies C99 so variable length array is available.
Differential Revision: https://phab.mercurial-scm.org/D1125
We got `alias_*_blocked` and `unknown_system_*` logged which bloated the
table. Since we are not using them, let's just remove them from logging.
Differential Revision: https://phab.mercurial-scm.org/D1064
This way we could get a handle on how many packs users are contending with.
Note that this location is unsatisfactory because we have so little metadata
available -- it'd be nice to denote type of store (data, history), type of
content (data, trees), and location of store (local, cache). Perhaps callers
could pass these keys to the constructor.
Differential Revision: https://phab.mercurial-scm.org/D1280
Summary:
Use python instead of jq in pullbackup test
fixes breakage reported in D6186148
Test Plan:
Run unit tests
Verify build runs without a hitch
Reviewers: swhitaker
Differential Revision: https://phabricator.intern.facebook.com/D6208705
This patch adds support to fold for outputting node changes as a part of it's
json output.
The nodechanges is a dictionary having predecessor as key and list of successors
as value. The successor is a list as we can have multiple successor for a
predecessor in case of split and having a generic format will be helpful.
After this patch, if you want json output from fold, do
`hg fold ..args.. -Tjson -q`
To just get the dictionary of hash changes, do
`hg fold ..args.. -T '{nodechanges|json}' -q`
-q is required to prevent the ui.status output coming from hg.update and rebase
part.
Since we have added support to fold for nodechanges, it has been added to
blacklist for cleanupnodes wrapper in tweakdefaults.
Differential Revision: https://phab.mercurial-scm.org/D1041
The logic of pushvars extension was moved to core and the extension here just
used to set a config which defaults to False in core. Let's move that config
setting to tweakdefaults and drop the extension.
Differential Revision: https://phab.mercurial-scm.org/D1275
Summary:
Instead of making an autobackup, make hg pullbackup simply
write the backup state from the recently restored backup.
Test Plan:
cd ~/facebook-hg-rpms/fb-hgext/tests
source ../../hg-dev
rt test-infinitepush-*.t
Reviewers: #mercurial, cdown, stash
Reviewed By: stash
Subscribers: mjpieters, medson
Differential Revision: https://phabricator.intern.facebook.com/D6186148
Tasks: T22856338
Tags: bootcamp
Signature: 6186148:1509457565:e66cefc6e95d9b1cfb91e9bf8fa5c40b05e4fb82
pushrebase can't use the standard rev.manifest() or rev[filepath] accessors
because they aren't aware of treemanifests. Let's refactor our manifest and
filectx accessor patterns into a separate function and use it in the two places
that need manifests.
test-treemanifest-server.t was failing before this. One test was modified
because the does-it-have-the-tree check is now done earlier, before the lock is
taken, so we abort before the output is printed.
Differential Revision: https://phab.mercurial-scm.org/D1251
This test looked like it's been funky since it was introduced. It harded code
an exception from a bug. The bug was that we're storing bundle manifests in a
cache during pushrebase, and because those manifests are hybrid manifests they
haven't actually loaded from disk. Later we close the bundle, then after that we
try to access these cached manifests, which throws because they can't read from
the file.
The recent change to read the copied files from the manifest before the lock is
taken, forces these manifests to be in loaded, thus fixing the problem. So let's
update the test.
When checking the bundle contents against the revisions it is being rebased
over, include the sources of renames, as changes made in those files also
conflict with the bundle.
Differential Revision: https://phab.mercurial-scm.org/D1199
In the old version of this extension it was too difficult to determine how to
resume or abort the command that generated conflicts. Let's add that
information so consumers can programatically continue or abort.
Differential Revision: https://phab.mercurial-scm.org/D713
Once `mmap` succeeded, it's no longer necessary to keep the underlying fd
open. Therefore just close them as an attempt to reduce file descriptors
opened by the process.
This is similar to 592c614a259f, but on the pure python side.
Summary:
Stupid bug corrupted files with extra \r on windows.
From python manual:
>On Windows, 'b' appended to the mode opens the file in binary mode, so there are also modes like 'rb', 'wb', and 'r+b'. Python on Windows makes a distinction between text and binary files; the end-of-line characters in text files are automatically altered slightly when data is read or written. This behind-the-scenes modification to file data is fine for ASCII text files, but it’ll corrupt binary data like that in JPEG or EXE files. Be very careful to use binary mode when reading and writing such files. On Unix, it doesn’t hurt to append a 'b' to the mode, so you can use it platform-independently for all binary files.
Test Plan: the crdump tests should now pass on windows as well (didn't actually run them on Windows because I don't really know how)
Reviewers: #mercurial, rmcelroy, ikostia, medson
Reviewed By: medson
Subscribers: mjpieters, medson
Differential Revision: https://phabricator.intern.facebook.com/D6151931
Tasks: T23027803
Signature: 6151931:1508964509:65806961cbde26884cd45def17767135ceeac05f
In 272dd7d8a1bb1b2e93697b5b65183b08c8bf0853 upstream hg commit phase pushes
started to use separate bundle2 part. Because of that:
1) pushkey hook is not called when updating phases.
2) pushkey bundle2 part doesn't process phases anymore, 'phase-heads' part do
it instead.
It caused breakages in infinitepush, because phases handling is not skipped even during
infinitepush. This diff fixes it.
Test Plan:
Run infinitepush tests
Differential Revision: https://phab.mercurial-scm.org/D1225
Upstream has added devel warnings that require us to register all the configs
and remove the defaults that are specified at read time. This doesn't fix all
the cases, but is the start.
Differential Revision: https://phab.mercurial-scm.org/D1206
Upstream has renamed this config. Since the old config name was still an alias
for the new name, it was technically only breaking the test output where we
hardcoded the config name in the output. But I went ahead and updated every
place anyways.
Differential Revision: https://phab.mercurial-scm.org/D1209
Summary:
The background repack uses the incorrect executable while running the
repack command due to which there are no logs for the background repack. This
commit fixes that by specifying the correct executable.
Test Plan:
- Ran all the tests.
- Tested the output of `util.hgexecutable()` with system hg and chg, and
verified that it returns the correct path. As expected, `util.hgcmd()`
returns the corresponding path for hg and chg in this case.
- Did a `hg pull` on one of the repos and verified that it triggered a
background incremental repack.
Reviewers: #fbhgext, quark
Reviewed By: #fbhgext, quark
Differential Revision: https://phab.mercurial-scm.org/D1221
In 272dd7d8a1bb1b2e93697b5b65183b08c8bf0853 upstream hg commit phase pushes
started to use separate bundle2 part. Because of that:
1) pushkey hook is not called when updating phases.
2) pushkey bundle2 part doesn't process phases anymore, 'phase-heads' part do
it instead.
It caused breakages in pushrebase. This diff fixes it.
Test Plan:
Run tests test-pushrebase* and test-treemanifest*
Differential Revision: https://phab.mercurial-scm.org/D1204
Previously test-treemanifest-treeonly.t was hanging. Now it doesn't hang (but
still fails).
It obviously doesn't fix the real issue, but at least it doesn't make test
hang.
Test Plan:
Run test-treemanifest-treeonly.t
Differential Revision: https://phab.mercurial-scm.org/D1205
This is what happened to D1048 and D1050. Pushrebase should ideally check
the copy source in additional to modified files and block the push.
Differential Revision: https://phab.mercurial-scm.org/D1172
