Support for the --kill-nailguns option was inadvertently removed, this puts it back.
Review Request #1352 - Created Nov. 18, 2014 and submitted
| Information | |
|---|---|
| Eric Ayers | |
| pants | |
| zundel/nailgun-timeouts-on-ci | |
| 803 | |
| cb9122a... | |
| Reviewers | |
| pants-reviews | |
| benjyw, ity, patricklaw | |
The kill-nailguns flag was being read through old_options.cleanup_nailguns
Also updated the nailgun timeout logic to try and pinpoint transient errors in CI if it
happens again.This reverts the change at https://rbcommons.com/s/twitter/r/1345/ which is redundant with this one (and not as effective.)
CI Passed.
Patrick Lawson
lgtm
-
src/python/pants/java/nailgun_executor.py (Diff revision 1) -
I would really love to get this into config. We find ourselves pushing these limits somewhat regularly when CI machines get a spike of jobs.
-
src/python/pants/java/nailgun_executor.py (Diff revision 1) -
Might as well flip this over to use
.formatfor consistency with below.
Change Summary:
Have to keep old_options code in place and guard against new_options not being present due to setup_py support still being around.
Went crazy converting format strings in nailgun_executor.py
Commit: |
|
||||
|---|---|---|---|---|---|
Diff: |
Revision 2 (+41 -34) |
-
src/python/pants/java/nailgun_executor.py (Diff revision 2) -
Note: The overall timeout is the same (50 seconds), but I've lengthened the initial connect timeout to 10 seconds from 5 seconds.
Patrick and I discussed making this configurable a bit. We could fetch the config with Config.from_cache(), but what we really want is a way to reference the new options system.
Ship It!
Ship It!