Closed 6543 closed 2 weeks ago
-> not tested jet but should fix it
Attention: Patch coverage is 71.42857%
with 8 lines
in your changes missing coverage. Please review.
Project coverage is 26.55%. Comparing base (
b0f52b9
) to head (0e3e42b
). Report is 27 commits behind head on main.
Files with missing lines | Patch % | Lines |
---|---|---|
pipeline/backend/common/script_posix.go | 42.85% | 3 Missing and 1 partial :warning: |
pipeline/backend/common/script_win.go | 55.55% | 3 Missing and 1 partial :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
I did not test but I don't see why working directory be anyhow related to workspace or permissions (it could be of course relative but not limited to)
As said in my previous comment, it is about the mount point and its permissions which is different for subdirs on k8s.
I wouldn't have changed it (i.e. get rid of workDir
) otherwise in the previous/initial PR.
While I am not saying that this PR won't work per se (before I tested it), I am not so happy about an (early) approval here without any (practical) tests (given the impact and size of the previous PR) without waiting for at least an reply from the former PR creator (which in this case happens to be me but that doesn't matter overall). I.e, if I wouldn't have put a "please wait" just shortly after, this one would have likely already been merged.
On a related note: as this is all just because of directory:
: arguably this is a niche param which almost not used anywhere and not doing much else than cd
to a subdir (for often no reason). If this happens to cause general issues for the whole rootless approach, I'd be in strong favor of removing it.
On a related note: as this is all just because of
directory:
: arguably this is a niche param which almost not used anywhere and not doing much else thancd
to a subdir (for often no reason). If this happens to cause general issues for the whole rootless approach, I'd be in strong favor of removing it.
I strongly disagree on this, we use it quite often, especially since for plugins you can't specify commands option to cd subdir
otherwise plugin itself won't run and would not be secure anymore but there are many cases where for example reviewdog plugins where you need to run eslint only for subdir where frontend code is located and golangci-lint on directory where backend code is, it's also way more clean this way
As said in my previous comment, it is about the mount point and its permissions which is different for subdirs on k8s. I wouldn't have changed it (i.e. get rid of
workDir
) otherwise in the previous/initial PR.While I am not saying that this PR won't work per se (before I tested it), I am not so happy about an (early) approval here without any (practical) tests (given the impact and size of the previous PR) without waiting for at least an reply from the former PR creator (which in this case happens to be me but that doesn't matter overall). I.e, if I wouldn't have put a "please wait" just shortly after, this one would have likely already been merged.
I'm sorry for a bit rushed approval but initially code looked good code wise and as this is something that is pretty much regression for me so I approved it 😊
The direcotory is important for plugins yes :)
I'd like to see both working (rootles & dir) ... I tested docker backend witch works just fine with workdir and an non-root user.
For k8s I need to test things, is pull here is first and formost to discusst getting a good solution.
Tearing down https://woodpecker-ci-woodpecker-pr-4319.surge.sh
close #4314