Open sluongng opened 3 months ago
The nogo failure looks like it's easier to fix than ignore.
@fmeum is it though?
The fieldalignment
analyzer is quite noisy and is triggered on proto-generated code? 🤔
Im in favor of disabling it with a TODO.
@fmeum is it though?
The
fieldalignment
analyzer is quite noisy and is triggered on proto-generated code? 🤔Im in favor of disabling it with a TODO.
True, I missed that this was the generated code, not the test. In that case I agree that disabling it by default is the right move.
Yeah. The other option is to add a relevant nogo_config.json
file to exclude proto-generated files for that analyzer.
I don't think it's worth the noise it gonna generate for downstream users though.
Hmm, these dependencies upgrades turn out to be a huge PITA.
Especially, for the popular_repo
test, I think we should consider dropping x/tools
testing entirely. The repo is where gopls
is being actively developed and the recent code quality is not that high. v0.19.0 contains a nil pointer bug and a bunch of build/test targets in that repo are not gazelle/rules_go compatible.
I also had to patch Gazelle temporarily to avoid analyzer duplication. It should be a temp workaround until (a) we remove that in Gazelle, then (b) release a new Gazelle version, and then (c) upgrade the Gazelle version in rules_go.
Upgrade golang.org/x/tools to v0.18.0. Note that the latest v0.19.0 has a bug that we should avoid using.
Update the analyzers (passes) in TOOLS_NOGO and the ones automatically included when
vet = True
is configured. The vet analyzers are taken fromgo tool vet help
documentation.