Temporary reduce TravisCI jobs

Hi Folks, it is just an idea because our credits on TravisCI are free but not unlimited. Each PR and each new commit on such a PR do create 5 testing jobs (in fact running containers) eating our Travis credits. I propose to deactivate 4 jobs and just keep Python 3.10 on AMD64 as the one and only testing job until we prepare the next release. For myself I sometimes create PR's against my own forked repo to run TravisCI with my own credits just to be sure. What do you think? It could spare a lot of credits. Kind Christian

Hi, Christian! I agree, every commit pushed to a PR causes quite a cascade of Tavis jobs. To find obvious problems like Syntax errors, one platform is enough. We can do more thorough testing when we get closer to releasing. Cheers, Michael On Wed, Nov 09, 2022 at 10:20:32AM +0000, c.buhtz@posteo.jp wrote:

On Wed, 2022-11-09 at 10:20 +0000, c.buhtz@posteo.jp wrote:
Each PR and each new commit on such a PR do create 5 testing jobs (in fact running containers) eating our Travis credits.
You mean the test matrix configured in .travis.yml (2 archs * 5 python versions minus 5 excluded combinations)?
I propose to deactivate 4 jobs and just keep Python 3.10 on AMD64 as the one and only testing job until we prepare the next release.
Good idea. This "saves" credits but eg. I also would never have seen a failing unit test on just python 3.6 since it happens only on travis but not in my anaconda environment: https://github.com/bit-team/backintime/pull/1319 How about keeping these combos: python 3.6 on amd64 python 3.9 on ppc64le python 3.10 on amd64 This covers - three python versions from oldest supported to current - and two architectures with only three jobs and is enough for 833 builds (25K credits / (3 jobs * 10 credit)). This should be enough for more than a year and BiT is a wide-spreaded FOSS app so I guess travis will support it in the future with more credits.

Hi, Christian! I agree, every commit pushed to a PR causes quite a cascade of Tavis jobs. To find obvious problems like Syntax errors, one platform is enough. We can do more thorough testing when we get closer to releasing. Cheers, Michael On Wed, Nov 09, 2022 at 10:20:32AM +0000, c.buhtz@posteo.jp wrote:

On Wed, 2022-11-09 at 10:20 +0000, c.buhtz@posteo.jp wrote:
Each PR and each new commit on such a PR do create 5 testing jobs (in fact running containers) eating our Travis credits.
You mean the test matrix configured in .travis.yml (2 archs * 5 python versions minus 5 excluded combinations)?
I propose to deactivate 4 jobs and just keep Python 3.10 on AMD64 as the one and only testing job until we prepare the next release.
Good idea. This "saves" credits but eg. I also would never have seen a failing unit test on just python 3.6 since it happens only on travis but not in my anaconda environment: https://github.com/bit-team/backintime/pull/1319 How about keeping these combos: python 3.6 on amd64 python 3.9 on ppc64le python 3.10 on amd64 This covers - three python versions from oldest supported to current - and two architectures with only three jobs and is enough for 833 builds (25K credits / (3 jobs * 10 credit)). This should be enough for more than a year and BiT is a wide-spreaded FOSS app so I guess travis will support it in the future with more credits.
participants (3)
-
BiT dev
-
c.buhtz@posteo.jp
-
Michael Büker