| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This pulls in a minor fix for a bug discovered by telemetry where
typ could crash during process cleanup (crbug.com/437115). There
are a couple of unrelated changes for the python packaging metadata
that should not affect anything in Chromium.
R=dtu@chromium.org
Review URL: https://codereview.chromium.org/837923006
Cr-Commit-Position: refs/heads/master@{#312292}
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This reworks the way typ handles multiprocessing again, to make the API
(and the implementation) simpler for the common case.
TBR=dtu@chromium.org
BUG=402172
Review URL: https://codereview.chromium.org/664123005
Cr-Commit-Position: refs/heads/master@{#300599}
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
From typ commit df1f462ec978d47c7f092e86f4be3fafc7cc446f:
Rework typ to handle unimportable modules better.
multiprocessing on windows requires that the __main__ module
be importable in order to work properly. If developers are mostly
working and testing on unix, it's possible for them to not realize this
and attempt to call typ from a wrapper script that may not be
importable, and then get unexpected errors.
This change makes the typ entry points check if it is being called
from an unimportable __main__, and error out if so. Developers
must then specify several different ways to work around this by passing
the 'win_multiprocessing' flag to typ.main() or typ.Runner.main().
The flag takes one of the values specified in the typ.WinMultiprocessing
class:
ignore - The tests are not expected to run on Windows, so ignore
the fact that __main__ is not importable. (If the tests
are actually run on Windows, it will err out).
run_serially - Run serially (one at a time) on Windows, regardless
of how many cores are available (hence multiprocessing is
not used).
spawn - Typ will convert the args it was passed back into a command
line, and spawn off an invocation of itself on Windows to actually
execute the tests in parallel (Since typ itself is importable,
this works fine). This mode cannot be used in conjunction
with contexts or setup and teardown functions at this time.
force - This is like 'spawn', except that it forces the spawn even
on non-Windows platforms. This mostly exists for testing.
As part of this change, we also no longer expose spawn_main() as a public
entry point (main() handles spawning instead), and merge the cmdline.py
code directly into runner, as it didn't make sense to split things
across the two modules once runner had to do the importability checking.
We also add a host.call_inline() wrapper that is a dumb wrapper
around subprocess.call(); it does not do the output capturing that
host.call() does, and is used when spawning the typ subprocess if needed.
TBR=dtu@chromium.org
BUG=402172
Review URL: https://codereview.chromium.org/660133004
Cr-Commit-Position: refs/heads/master@{#300052}
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change pulls in the new --all flag for typ (so that we can run
disabled tests), and pulls in the ability to set default values for
the command line arguments.
It also adds a bunch more test coverage, but that testing didn't expose
any common bugs as I recall ...
TBR=dtu@chromium.org
BUG=402172
Review URL: https://codereview.chromium.org/654123003
Cr-Commit-Position: refs/heads/master@{#299630}
|
|
|
This change adds a new python testing framework that will replace
the various adhoc variants of this we use in Chromium and Blink,
at least temporarily until if/when we can switch to infra's
expect_tests framework.
Using this library will allow us to consolidate a bunch of duplicated
code used for telemetry_unittests, webkit_python_tests,
mojo_python_tests, webkit_tests, and elsewhere.
TBR=cpu@chromium.org, dtu@chromium.org
BUG=402172
Review URL: https://codereview.chromium.org/627763002
Cr-Commit-Position: refs/heads/master@{#299004}
|
