Hi,
Perhaps it would be possible to set up a Travis CI matrix entry for OS X, those builds are often quite slow but at least it could be part of the "allowed failures" suite. That would help detect platform issues at PR time rather than later :-)
Regards
Antoine.
On May 2, 2017, at 5:35 PM, Antoine Pitrou <antoine@python.org> wrote:
Hi,
Perhaps it would be possible to set up a Travis CI matrix entry for OS X, those builds are often quite slow but at least it could be part of the "allowed failures" suite. That would help detect platform issues at PR time rather than later :-)
I think the only reason we don’t have them on is because the macOS builds on Travis are _Super_ slow and regularly get a large backlog. Fast Finish and Allowed Failures would help with that though.
— Donald Stufft
Is there a way to get faster OS X builds with our connections at Travis-CI? I agree that OS X builds are usually very slow (though it depends on daytime, see https://www.traviscistatus.com/#week), but perhaps it's possible to improve on that :-)
Regards
Antoine.
Le 02/05/2017 à 23:37, Donald Stufft a écrit :
On May 2, 2017, at 5:35 PM, Antoine Pitrou <antoine@python.org <mailto:antoine@python.org>> wrote:
Hi,
Perhaps it would be possible to set up a Travis CI matrix entry for OS X, those builds are often quite slow but at least it could be part of the "allowed failures" suite. That would help detect platform issues at PR time rather than later :-)
I think the only reason we don’t have them on is because the macOS builds on Travis are _Super_ slow and regularly get a large backlog. Fast Finish and Allowed Failures would help with that though.
— Donald Stufft
Travis's macOS builds aren't as slow as they used to be, between them adding capacity and our queue increase.
Alex
On Tue, May 2, 2017 at 5:38 PM, Antoine Pitrou <antoine@python.org> wrote:
Is there a way to get faster OS X builds with our connections at Travis-CI? I agree that OS X builds are usually very slow (though it depends on daytime, see https://www.traviscistatus.com/#week), but perhaps it's possible to improve on that :-)
Regards
Antoine.
Le 02/05/2017 à 23:37, Donald Stufft a écrit :
On May 2, 2017, at 5:35 PM, Antoine Pitrou <antoine@python.org <mailto:antoine@python.org>> wrote:
Hi,
Perhaps it would be possible to set up a Travis CI matrix entry for OS X, those builds are often quite slow but at least it could be part of the "allowed failures" suite. That would help detect platform issues at PR time rather than later :-)
I think the only reason we don’t have them on is because the macOS builds on Travis are _Super_ slow and regularly get a large backlog. Fast Finish and Allowed Failures would help with that though.
— Donald Stufft
python-committers mailing list python-committers@python.org https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
-- "I disapprove of what you say, but I will defend to the death your right to say it." -- Evelyn Beatrice Hall (summarizing Voltaire) "The people's good is the highest law." -- Cicero GPG Key fingerprint: D1B3 ADC0 E023 8CA6
So Donald is right that the reason we don't have macOS builds through Travis is due to the fact that back in February they had a horrible backlog and added too much time to the total build.
But as Alex pointed out, Travis has subsequently increased their macOS fleet and given the Python organization a better capacity (it should be mentioned all we can get from travis without money is the ability to run more concurrent jobs, but that doesn't solve the problem if the macOS fleet in Travis is itself the bottleneck).
If someone wants to author a PR we could test the time on the master branch and see what tweaks (if any) make the time added acceptable to everyone.
On Tue, 2 May 2017 at 16:43 Alex Gaynor <alex.gaynor@gmail.com> wrote:
Travis's macOS builds aren't as slow as they used to be, between them adding capacity and our queue increase.
Alex
On Tue, May 2, 2017 at 5:38 PM, Antoine Pitrou <antoine@python.org> wrote:
Is there a way to get faster OS X builds with our connections at Travis-CI? I agree that OS X builds are usually very slow (though it depends on daytime, see https://www.traviscistatus.com/#week), but perhaps it's possible to improve on that :-)
Regards
Antoine.
Le 02/05/2017 à 23:37, Donald Stufft a écrit :
On May 2, 2017, at 5:35 PM, Antoine Pitrou <antoine@python.org <mailto:antoine@python.org>> wrote:
Hi,
Perhaps it would be possible to set up a Travis CI matrix entry for OS X, those builds are often quite slow but at least it could be part of the "allowed failures" suite. That would help detect platform issues
at
PR time rather than later :-)
I think the only reason we don’t have them on is because the macOS builds on Travis are _Super_ slow and regularly get a large backlog. Fast Finish and Allowed Failures would help with that though.
— Donald Stufft
python-committers mailing list python-committers@python.org https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
-- "I disapprove of what you say, but I will defend to the death your right to say it." -- Evelyn Beatrice Hall (summarizing Voltaire) "The people's good is the highest law." -- Cicero GPG Key fingerprint: D1B3 ADC0 E023 8CA6
python-committers mailing list python-committers@python.org https://mail.python.org/mailman/listinfo/python-committers Code of Conduct: https://www.python.org/psf/codeofconduct/
2017-05-02 23:37 GMT+02:00 Donald Stufft <donald@stufft.io>:
I think the only reason we don’t have them on is because the macOS builds on Travis are _Super_ slow and regularly get a large backlog. Fast Finish and Allowed Failures would help with that though.
Maybe we can start with a small subset of tests and enlarge it later?
I may help to catch the most obvious failures.
Sadly, in my experience, on macOS, I only saw subtle failures, no obvious bug in very common tests.
Victor
participants (5)
-
Alex Gaynor -
Antoine Pitrou -
Brett Cannon -
Donald Stufft -
Victor Stinner