Opened 10 years ago
Closed 6 years ago
#3398 defect closed fixed (fixed)
process cleanup for ReactorBuilder-based tests loops indefinitely if processes do not exit
| Reported by: | zooko | Owned by: | Itamar Turner-Trauring |
|---|---|---|---|
| Priority: | normal | Milestone: | |
| Component: | core | Keywords: | |
| Cc: | Jean-Paul Calderone | Branch: |
branches/process-cleanup-loop-3398-2
branch-diff, diff-cov, branch-cov, buildbot |
| Author: | itamarst, exarkun |
Description
http://buildbot.twistedmatrix.com/builders/cygwin-py2.5-select/builds/1
Produces many hundreds or thousands of this line in the test.log:
2008-08-26 15:29:02-0700 [-] ReactorBuilder.tearDown reaping some processes {2540: <PTYProcess pid=2540 status=-1>}
Followed by many of this:
2008-08-26 15:29:10-0700 [-] ReactorBuilder.tearDown reaping some processes {2524: <PTYProcess pid=2524 status=-1>}
and eventually I stopped it.
Change History (12)
comment:1 Changed 10 years ago by
| Cc: | Jean-Paul Calderone added |
|---|
comment:3 Changed 10 years ago by
Maybe. I don't have cygwin installed anywhere, so I can't test this.
comment:5 Changed 10 years ago by
| Resolution: | duplicate |
|---|---|
| Status: | closed → reopened |
Even if cygwin is broken, the tests shouldn't get stuck in a loop. #3529 can be for fixing cygwin support somehow. I'm using this ticket to fix the getting stuck behavior.
comment:6 Changed 10 years ago by
| Owner: | changed from Glyph to Jean-Paul Calderone |
|---|---|
| Status: | reopened → new |
comment:7 Changed 10 years ago by
| Author: | → exarkun |
|---|---|
| Branch: | → branches/process-cleanup-loop-3398 |
(In [25347]) Branching to 'process-cleanup-loop-3398'
comment:8 Changed 10 years ago by
| Owner: | changed from Jean-Paul Calderone to zooko |
|---|---|
| Summary: | cygwin unit tests stuck in loop → process cleanup for ReactorBuilder-based tests loops indefinitely if processes do not exit |
Can you try the branch on cygwin?
I expect the tests will fail, but the suite should run to completion at least.
comment:9 Changed 6 years ago by
| Owner: | changed from zooko to Itamar Turner-Trauring |
|---|
comment:10 Changed 6 years ago by
| Keywords: | cygwin removed |
|---|
This is happening on non-cygwin platforms as well, e.g.: http://buildbot.twistedmatrix.com/builders/fedora32-py2.5-reactors/builds/1553/steps/poll/logs/test.log
comment:11 Changed 6 years ago by
| Author: | exarkun → itamarst, exarkun |
|---|---|
| Branch: | branches/process-cleanup-loop-3398 → branches/process-cleanup-loop-3398-2 |
(In [34151]) Branching to 'process-cleanup-loop-3398-2'
comment:12 Changed 6 years ago by
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
It seems the process started by
twisted.internet.test.test_process.ProcessTestsBuilder_PollReactor.test_childConnectionLostnever exits. It is supposed to exit when it receives a blank line on stdin, which the test sends it after sending two other lines.It seems like the test method actually ran to completion and all the events it expected did happen because the infinite log message happens in
tearDownwhich should only run after the test method has returned.So the question here may be about why the child process isn't exiting when it is expected to.