This repository was archived by the owner on Apr 23, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Apr 23, 2026. It is now read-only.
rules_cc dependency marked as dev causes rules_cc module visibility issues #203
Description
When trying to use the following code:
load("@rules_proto//proto:defs.bzl", "proto_descriptor_set")
proto_descriptor_set(
name = "my_fdset",
deps = ["...."],
)This causes the following error:
ERROR: error loading package '@@rules_proto~6.0.0-rc2//tools/file_concat': Unable to find package for @@[unknown repo 'rules_cc' requested from @@rules_proto~6.0.0-rc2]//cc:defs.bzl: The repository '@@[unknown repo 'rules_cc' requested from @@rules_proto~6.0.0-rc2]' could not be resolved: No repository visible as '@rules_cc' from repository '@@rules_proto~6.0.0-rc2'.
Reacted by Spencer Connaughton and Saleh
Activity
comius commentedon Feb 9, 2024
cc @alexeagle @thesayyn, please triage/fix
Introduce the concept of known EventGroup metadata types. (#1512)
alexeagle commentedon Apr 22, 2024
Yup, that tool is written in C++ which is naughty since we only declare that as a devDep:
rules_proto/MODULE.bazel
Line 16 in f889a1b
and many users have broken C++ toolchains.
The program itself is very small https://github.com/bazelbuild/rules_proto/blob/f889a1b532fdca5f5051691f023a6a9f37ce494f/tools/file_concat/main.cc
I think we should avoid having tools live in rules_proto.
SanjayVas commentedon May 1, 2024
It looks like despite being reported against v6.0.0-rc2, this bug continued into both v6.0.0-rc3 and the final v6.0.0 release. Can we get a fix either in rules_proto itself (v6.0.1) or in BCR (6.0.0.bcr.1)?
alexeagle commentedon May 1, 2024
What do you propose as the fix? I suppose our only choice per semver is to make rules_cc a non-dev dependency, rather than remove this tool or ship pre-built binaries for it.
SanjayVas commentedon May 1, 2024
Yes, exactly this for the short term as right now the module definition is just incorrect/broken. Anything like what was proposed in #203 (comment) would need to come later.
Introduce the concept of known EventGroup metadata types. (#1512)