Nicklas Larsson

> I made a test: > > ``` > cd raster/r.mapcalc/ > make clean > make -j21 > make -j21 > make -j21 > ``` > > Only at the...

> So I re-ran this entirely and the errors were the same. Interesting. Did this happen before this PR (ie. before and after #3415 was merged)?

@wjcper2008 Will adding the flag `-m` (disk swap memory option) to the `r.watershed` command make any difference?

For reference: QGIS make use of https://github.com/qgis-bot, for some automatic procedures.

~~Related: https://github.com/OSGeo/grass/issues/1125~~ Rereading the description, now I see this is about interface as in API, not GUI. Sorry about the noise.

We may also remove this workflow, replace it with the use of `pre-commit.ci`, see #3255, which not only addresses this but also a number of other issues.

> Is that PR/configuration almost ready? The GRASS repo is ready now, it only remains to activate it. The advantage is that all pre-commit hooks defined in https://github.com/OSGeo/grass/blob/main/.pre-commit-config.yaml is run...

> And would the checks be executed on forks, ie, would be able to run on a PR inside my own fork, to make sure everything is ready before submitting...

The number of "accepted" failures for win is 60 per cent, set with: https://github.com/OSGeo/grass/blob/e4431cc79acf7f5ca42389c21e100f3c9bc51c56/.github/workflows/test_thorough.bat#L5 This is obviously not good, the reason to not remove the failing tests is to not...

You can, at least on your testing fork, add a config file to exclude specific failing tests for windows, like for the Mac: https://github.com/OSGeo/grass/blob/e4431cc79acf7f5ca42389c21e100f3c9bc51c56/.github/workflows/macos.yml#L74-L77