Fabian Meumertzheim
Fabian Meumertzheim
> I am afraid that this is a way to pop up internally now (in fact, quite a lot; to make matters worse, in that case it is benign), so...
@sgowroji I think that this would fit the docs team better.
@fweikert Would you be available to review this PR?
@tjgq Could you take a quick glance and let me know whether you approve of the general approach (introducing `TreeEmptyDirectoryArtifact` to complement `TreeFileArtifact` as a subclass of the new `TreeChildArtifact`)?...
Please note that there are also special cases I haven't dealt with yet outside of test code: For example, the Starlark `DirectoryExpander` should probably filter out `TreeEmptyDirectoryArtifact`s.
Let me first clarify the intention behind this PR: I view #15789 and #15901 as serious correctness issues in a supported feature that is enabled by default (directory outputs). The...
I agree that if we decide that `TreeArtifact`s are merely collections of files, we would either have to delete empty subdirectories after the tree has been generated or, more drastically,...
@sgowroji I think that this issue is still valid. It isn't tied to a particular project, it's just about better error messages. @gregestren who I think has an ongoing project...
@codeman9 If you really need to work around this, you could probably define an `alias` in a non-main repository pointing to the file and then provide the label of the...
@bazel-io flag