This patch adds a global lint that only runs when a file or directory
that matches their configuration (via `extensions` and `exclude`) has
been modified or specified.
Global lints never shard into chunks; they are, by definition global
(i.e., across the entire source tree) and act on all inputs in a
single invocation. It's up to the global lint to manage command line
sizes, etc. Since batching is handled by the lint type but sharding
is handled by the lint roller, there's a little abstraction leak so
that the lint type can control how its invocation is sharded: the
existing `batch` member is generalized from the existing `True` and
`False` to add a new `"global"` value which disables sharding.
Differential Revision: https://phabricator.services.mozilla.com/D35275
--HG--
extra : moz-landing-system : lando
The following python-test paths are being moved out of 'make check' and into their own task:
- python/mozlint
- testing/mozbase
- tools/lint
The following python-test paths previously did not run on Windows:
- python/mozterm
- testing/marionette
- testing/raptor
- tools/tryselect
Differential Revision: https://phabricator.services.mozilla.com/D10759
--HG--
extra : moz-landing-system : lando
The following python-test paths are being moved out of 'make check' and into their own task:
- python/mozlint
- testing/mozbase
- tools/lint
The following python-test paths previously did not run on Windows:
- python/mozterm
- testing/marionette
- testing/raptor
- tools/tryselect
MozReview-Commit-ID: C07FANaYzf7
Depends on D10758
Differential Revision: https://phabricator.services.mozilla.com/D10759
--HG--
extra : moz-landing-system : lando
This assumes that a lineno of 0 denotes a "file-level" issue, e.g an issue
associated with the filename. In the future it might be better to treat these
"file-level" issues as first class citizens. This would involve updating things
like formatters, editor integrations and review bot to not assume a lineno.
Differential Revision: https://phabricator.services.mozilla.com/D10383
--HG--
extra : moz-landing-system : lando
We were currently adding exclude paths to the "discard" set if the path contains
the include, but we weren't adding them if the include contains the path.
Depends on D5863
Differential Revision: https://phabricator.services.mozilla.com/D8845
--HG--
extra : moz-landing-system : lando
Right now there are excludes defined in the linter definition (via the .yml
files), as well as excludes defined in lintargs (via the mach_commands.py).
This is a minor simplification that extends each linter definition's local
excludes with the global ones right off the bat. This just makes it a bit
easier to keep track of.
Depends on D5863
Differential Revision: https://phabricator.services.mozilla.com/D8844
--HG--
extra : moz-landing-system : lando
Currently pathutils.filterpaths computes some extra paths that the underlying linter
should try to exclude if it can. It sets these on lintargs['exclude'], and relies on
the fact that it was passed by reference to work.
However, it seems like pytest does some magic under the hood that prevents this value
from being propagated back. It will also fail if 'filterpaths' was invoked in a
subprocess.
It's much simpler and cleaner to pass this value back directly, rather than rely on a
reference.
Differential Revision: https://phabricator.services.mozilla.com/D8850
--HG--
extra : moz-landing-system : lando
I missed an edge case where the computed base itself was specified in the
paths input.
Differential Revision: https://phabricator.services.mozilla.com/D8842
--HG--
extra : moz-landing-system : lando
Often we specify globs in our exclude patterns, e.g:
exclude:
- **/node_modules
- obj*
However, these globs get expanded out to *every* file that matches them. This
can sometimes be thousands or even tens of thousands of files.
We then pass these paths on to the underlying linters and tell them to
exclude them all. This causes a lot of overhead and slows down performance.
This commit implements a "collapse" function. Given a set of paths, it'll
collapse them into the smallest set of parent directories that contain the
original set, and that don't contain any extra files.
For example, given a directory structure like this:
a
-- foo.txt
-- b
-- bar.txt
-- baz.txt
-- c
-- ham.txt
-- d
-- spam.txt
Then the following will happen:
>>> collapse(['a/foo.txt', 'a/b/bar.txt', 'a/c/ham.txt', 'a/c/d/spam.txt'])
['a/foo.txt', 'b/bar.txt', 'c']
Since all files under directory 'c' are specified by the original set (both
'c/ham.txt' and 'c/d/spam.txt'), we can collapse it down to just 'c'. However
not all files under 'b' are specified (we're missing 'a/b/baz.txt'), so we
can't collapse 'b' (and by extension also can't collapse 'a').
If we had included 'a/b/baz.txt':
>>> collapse(['a/foo.txt', 'a/b/bar.txt', 'a/b/baz.txt', 'a/c/ham.txt', 'a/c/d/spam.txt'])
['a']
In both cases, the smallest set of paths that contains the original set (and
only the original set) is computed.
The collapse function has a little bit of overhead but it's not too bad.
For example collapsing all files matched by '**/node_modules' takes ~0.015s.
Collapsing two full objdirs, takes ~0.6s. But a follow up commit is planned to
make sure we stop using 'obj*' to reduce that overhead.
Depends on D7738
Differential Revision: https://phabricator.services.mozilla.com/D7739
--HG--
extra : moz-landing-system : lando
Often we specify globs in our exclude patterns, e.g:
exclude:
- **/node_modules
- obj*
However, these globs get expanded out to *every* file that matches them. This
can sometimes be thousands or even tens of thousands of files.
We then pass these paths on to the underlying linters and tell them to
exclude them all. This causes a lot of overhead and slows down performance.
This commit implements a "collapse" function. Given a set of paths, it'll
collapse them into the smallest set of parent directories that contain the
original set, and that don't contain any extra files.
For example, given a directory structure like this:
a
-- foo.txt
-- b
-- bar.txt
-- baz.txt
-- c
-- ham.txt
-- d
-- spam.txt
Then the following will happen:
>>> collapse(['a/foo.txt', 'a/b/bar.txt', 'a/c/ham.txt', 'a/c/d/spam.txt'])
['a/foo.txt', 'b/bar.txt', 'c']
Since all files under directory 'c' are specified by the original set (both
'c/ham.txt' and 'c/d/spam.txt'), we can collapse it down to just 'c'. However
not all files under 'b' are specified (we're missing 'a/b/baz.txt'), so we
can't collapse 'b' (and by extension also can't collapse 'a').
If we had included 'a/b/baz.txt':
>>> collapse(['a/foo.txt', 'a/b/bar.txt', 'a/b/baz.txt', 'a/c/ham.txt', 'a/c/d/spam.txt'])
['a']
In both cases, the smallest set of paths that contains the original set (and
only the original set) is computed.
The collapse function has a little bit of overhead but it's not too bad.
For example collapsing all files matched by '**/node_modules' takes ~0.015s.
Collapsing two full objdirs, takes ~0.6s. But a follow up commit is planned to
make sure we stop using 'obj*' to reduce that overhead.
Depends on D7738
Differential Revision: https://phabricator.services.mozilla.com/D7739
--HG--
extra : moz-landing-system : lando
There is only a single linter (test-disable.yml) that uses a glob in any
include path, and that usage is easily replaced by using the 'extensions' key
instead.
Since globs in include directives aren't very useful, let's disallow them. This
will allow us to simplify the 'filterpaths' logic quite substantially and make
future refactorings in this area easier.
Differential Revision: https://phabricator.services.mozilla.com/D6798
--HG--
extra : moz-landing-system : lando
This makes things more explicit. Previously we were relying on those magic
global "linters" variables, and it turned out that one of the tests was
actually linting a completely different set of linters than I was expecting.
This changes things so each test needs to explicitly define which linters it
wants to use.
Depends on D6410
Differential Revision: https://phabricator.services.mozilla.com/D6412
--HG--
extra : moz-landing-system : lando
This makes this test match all the other tests (which are named after the module
they are testing).
Also rename the test function to 'test_filterpaths'.
Differential Revision: https://phabricator.services.mozilla.com/D6410
--HG--
rename : python/mozlint/test/test_filterpaths.py => python/mozlint/test/test_pathutils.py
extra : moz-landing-system : lando
This will make it easier to add new test cases in the future. It'll
also make it easier to print debug, as only the output for each
specific test case will be printed on failure.
Differential Revision: https://phabricator.services.mozilla.com/D5862
--HG--
extra : moz-landing-system : lando
After fixing the absolute path issue in codespell, I noticed that the stylish
formatter doesn't indent lint issues that don't have a column properly. This
was never noticed before since most other linters have a column attribute.
Depends on D4012
Differential Revision: https://phabricator.services.mozilla.com/D4013
--HG--
extra : moz-landing-system : lando
As of this patch, any lint issue at the "warning" level will *only* be displayed
if --warnings is passed in. This gives us some flexibility to track issues that
are "recommended to fix" but that aren't required (aka don't cause a backout).
I think it would be ideal if the reviewbot ran with warnings enabled, and CI
ran without warnings. This way these issues will be surfaced to developers
(and hopefully get fixed prior to landing), but developers will always be able
to ignore them if the situation warrants it.
Since the last change converted everything to use errors, this patch should
be a no-op for now. However as we move forward (and potentially decide that
certain types of lint issues should result in a backout), this feature will
start seeing more and more use.
Depends on D3820
Differential Revision: https://phabricator.services.mozilla.com/D3821
--HG--
extra : moz-landing-system : lando
Currently there are 3 things that can impact the result of a lint run:
1. The list of lint issues found
2. The set of failures that happened during the setup phase
3. The set of failures that happened during the execution phase
All three of these things are stored as instance variables on the LintRoller
object, and then passed into a formatter when it comes time to print the
results. I'd like to add even more things that can impact the result, and it
became clear that the current scenario does not scale well.
This patch moves all data that could impact the end result of a lint run off of
the LintRoller object and onto a new 'result.ResultSummary' class. To avoid
confusion, this patch also renames the 'result.ResultContainer' class to
'result.Issue'.
With this new nomenclature:
result -> overall state of an entire lint run (can comprise multiple linters)
issue -> one specific lint infraction (at either 'warning' or 'error' level)
failure -> a non-recoverable error in the linter implementation itself
A "result" is comprised of 0 or more "issues" and 0 or more "failures".
Differential Revision: https://phabricator.services.mozilla.com/D3819
--HG--
extra : moz-landing-system : lando
This will make sure that when running |mach python-test --python 3| locally,
we only run the tests that also run in CI with python 3 (and therefore pass
presumably).
MozReview-Commit-ID: 3OBr9yLSlSq
--HG--
extra : rebase_source : 456340d0ecdddf1078f2b5b4ebb1eddf3813b26a
Files returned from version control (i.e via --outgoing or --workdir), are currently joined to
cwd. This will cause failures if |mach lint| is run from anywhere other than topsrcdir.
However we *do* want to join manually specified paths to cwd so things like:
cd devtools && mach lint client
continue to work. This patch makes sure we join the proper kind of path to the proper place.
MozReview-Commit-ID: EQmRhAr3Oog
--HG--
extra : rebase_source : 2629cc27f79059e44369d46d4f8278f83923582c
Previously, using --workdir or --outgoing could miss errors when modifying a
support file since those could affect unmodified files.
This patch allows linters to define support-files in their .yml configuration.
If using --outgoing or --workdir and a file matching one of those patterns was
modified, we'll lint the entire tree.
MozReview-Commit-ID: CuGLYwQwiWr
--HG--
extra : rebase_source : 00d4107c41404f5e6ab05e0106d5cd377e25652f
Since I left the next two patches to bitrot, I realized that a path I had added
didn't exist anymore. We should definitely error out if non-existant paths are
specified, otherwise the lists will become outdated and it will be possible to
accidentally disable linting on some files.
I discovered a few instances of this already in our existing definitions.
MozReview-Commit-ID: 8jsTKLI0nFE
--HG--
extra : rebase_source : acceb0b129fc472fb456ff527e4c8c52228edd59
There a few pieces needed here to properly handle KeyboardInterrupts.
1. All in-progress work needs to abort. Ideally the underlying linters will be
able to catch KeyboardInterrupt, and return partial results (like the flake8
linter does). Linters may alternatively allow the KeyboardInterrupt to
propagate up. Mozlint will catch and handle this appropriately, though any
results found will be lost. The only change to this behaviour was fixing a bug
in the flake8 linter.
2. Any unstarted jobs need to be canceled. In concurrent.futures, there are two
different queues. First, jobs are placed on the work queue, which is just a list
maintained by the parent process. As workers become available, jobs are moved
off the work queue, and onto the call queue (which is a multiprocessing.Queue).
Jobs that live on the work queue can be canceled with 'future.cancel()', whereas
jobs that live on the call queue cannot. The number of extra jobs that are stored
on the call queue is determined by this variable:
https://hg.mozilla.org/mozilla-central/file/deb7714a7bcd/third_party/python/futures/concurrent/futures/process.py#l86
In this patch, the parent process' sigint handler (which will be called on Ctrl-C)
is responsible for canceling all the jobs on the work queue. For the jobs on the
call queue, the best we can do is set a global variable that tells workers to
abort early.
3. Idle workers should exit gracefully. When there are no more jobs left, workers
will block on the call queue (either waiting for more jobs, or waiting for the
executor to send the shutdown signal). If a KeyboardInterrupt is received while a
worker is blocking, it isn't possible to intercept that anywhere (due to quirks
of how concurrent.futures is implemented). The InterruptableQueue class was
created to solve this problem. It will return None instead of propagating
KeyboardInterrupt. A None value will wake the worker up and tell it to gracefully
shutdown. This way, we avoid cryptic tracebacks in the output.
With all of these various pieces solved, pressing Ctrl-C appears to always exit
gracefully, sometimes even printing partial results.
MozReview-Commit-ID: 36Pe3bbUKmk
--HG--
extra : rebase_source : d4c312ee5cc3679eeee1407c5521aed679f84ad4
extra : source : a93a00141bf62f6bc9e30934c0e56f6b2e434bf0
This commit doesn't change any behaviour, just attempts to make this a little
more readable. The workers will call '_collect_results' for each WorkItem they
process (either because it is finished or because it was canceled).
This also differentiates between setup failures and run failures.
MozReview-Commit-ID: 36Pe3bbUKmk
--HG--
extra : rebase_source : 873167512b745ccdc52de7e7f1ecf66b094e063d
This formatter is useful for triaging paths when enabling new linters or
expanding existing ones. It works well with the -n/--no-filter option.
For example, if I wanted to look for candidates of new directories to enable
the codespell linter on, I could run:
./mach lint -l codespell -nf summary
This will print something like:
accessible: 429
dom: 142
layout: 15
testing: 53
etc..
If desired, you can also specify a depth by setting MOZLINT_SUMMARY_DEPTH. A
depth of 2 means results will be aggregated further down, e.g:
accesible/build: 129
accesible/ipc: 300
dom/indexedDB: 100
dom/workers: 42
etc..
The depth is always relative to the common path prefix of all results, so
running:
./mach lint -l codespell -nf python/mozbuild
Would expand all the directories under python/mozbuild (not topsrdir).
MozReview-Commit-ID: OiihLTpULA
--HG--
extra : rebase_source : eaaabc1d5cdc8e3942808d01b24e22081fea752e
This allows linters to define a 'setup' method which will automatically be
called by |mach lint| before running the linter. Users can also explicitly run
these methods (without doing any actual linting) by running |mach lint --setup|.
MozReview-Commit-ID: 74aY1pfsaX1
--HG--
extra : rebase_source : e6a7d769ba14c996c7a77316928063fa46358c5a
The initial motivation for this patch, was to prevent command lines that are
too long on Windows. To that end, there is a cap to the number of paths that
can be run per job. For now that cap is set to 50. This will allow for an
average path length of 160 characters, which should be sufficient with room to
spare.
But another big benefit of this patch is that we are running more things in
parallel. Previously, mozlint ran each linter in its own subprocess, but that's
it. If running eslint is 90% of the work, it'll still only get a single
process. This means we are wasting cores as soon as the other linters are
finished.
This patch chunks the number of specified paths such that there will be N*L
jobs where 'N' is the number of cores and 'L' is the number of linters. This
means even when there's a dominant linter, we'll be making better use of our
resources. This isn't perfect of course, as some paths might contain a small
number of files, and some will contain a very large number of files. But it's
a start
A limitation to this approach is that if there are fewer paths specified than
there are cores, we won't schedule enough jobs per linter to use those extra
cores. One idea might be to expand specified directories and individually list
all the paths under the directory. But this has some hairy edge cases that
would be tough to catch. Doing this in a non-hacky way would also require a
medium scale refactor.
So I propose further parallelization efforts be destined for follow-ups.
MozReview-Commit-ID: JRRu13AFaii
--HG--
extra : rebase_source : 242fb71fe0af8bd2a981bd10a7216bb897fe00ac
The initial motivation for this patch, was to prevent command lines that are
too long on Windows. To that end, there is a cap to the number of paths that
can be run per job. For now that cap is set to 50. This will allow for an
average path length of 160 characters, which should be sufficient with room to
spare.
But another big benefit of this patch is that we are running more things in
parallel. Previously, mozlint ran each linter in its own subprocess, but that's
it. If running eslint is 90% of the work, it'll still only get a single
process. This means we are wasting cores as soon as the other linters are
finished.
This patch chunks the number of specified paths such that there will be N*L
jobs where 'N' is the number of cores and 'L' is the number of linters. This
means even when there's a dominant linter, we'll be making better use of our
resources. This isn't perfect of course, as some paths might contain a small
number of files, and some will contain a very large number of files. But it's
a start
A limitation to this approach is that if there are fewer paths specified than
there are cores, we won't schedule enough jobs per linter to use those extra
cores. One idea might be to expand specified directories and individually list
all the paths under the directory. But this has some hairy edge cases that
would be tough to catch. Doing this in a non-hacky way would also require a
medium scale refactor.
So I propose further parallelization efforts be destined for follow-ups.
MozReview-Commit-ID: JRRu13AFaii
--HG--
extra : rebase_source : 6cd73d8b6888723de3410df043f7ed042ba3349f
This patch introduces a new report formatter for the mozlint
framework used by "./mach lint" that respects Unix output conventions,
popularised by grep(1), compilers, and preprocessors.
The output format looks like this:
testing/marionette/driver.js:1153:48: no-unused-vars error: 'resp' is defined but never used.
Many editors (ed, Emacs, vi, Acme) recognise this format, allowing
users to interact with the output like a hyperlink to jump to the
specified location in a file.
MozReview-Commit-ID: 3IyiFmNbtMY
--HG--
extra : rebase_source : a2628a999c187a1c43f3cc5d32e6db835028eca4
This also migrates the vcs.py test to mozversioncontrol and adds a new task for
it.
MozReview-Commit-ID: 9jTRkjNupVA
--HG--
extra : rebase_source : 400f27498e00ea45234ad7c951770b098e916b8e
Currently line linters (linters that open a file and process it line by line,
by applying a regex for example), don't handle directories. If a directory is
passed in, it will try to 'open' it, which fails. Directories can get hit if
the linter has a directory in its include directive or if the user passes in
--no-filter.
This patch modifies LineLinters so that if a directory is detected, we search
for all relevant files under that directory. If 'extensions' is used, we'll
look for only files with appropriate extensions. Otherwise we assume the
linter wants every file.
MozReview-Commit-ID: D9lzTNuQTob
--HG--
extra : rebase_source : 0b952c06eae28b67b687813ff7e75b231b2dd4d3
This switches most tests over to use pytest as the runner instead of unittest (taking
advantage of the fact that pytest can run unittest based tests).
There were a couple tests that had failures when swithing to pytest:
config/tests/unit-expandlibs.py
xpcom/idl-parser/xpidl/runtests.py
For these tests, I added a runwith='unittest' argument so that they still run the
same way as before. Once we fix them to use pytest, the unittest logic in mozunit.py
can be deleted.
MozReview-Commit-ID: Gcsz6z8MeOi
--HG--
extra : rebase_source : 3c762422ce0af54cbbe7d9fc20085a2d1ebe7057
Nick Alexander
Bastien Abadie
Sylvestre Ledru
Robert Bartlensky
Axel Hecht
Andrew Halberstadt
Andreea Pavel
Csoregi Natalia
Gurzau Raul
Andreas Tolfsen
Steve Armand
Tom Prince