dblaikie added a comment. In D116313#3222508 <https://reviews.llvm.org/D116313#3222508>, @hans wrote:
> In D116313#3218224 <https://reviews.llvm.org/D116313#3218224>, @aganea wrote: > >> I was wondering, can we mandate `LLVM_ENABLE_WERROR=ON` for all bots at >> least? (or maybe enable it through cmake, if ever there's a setting to >> detect we're running as a bot) >> Or just reverse the default setting on the main branch, ie. make >> `LLVM_ENABLE_WERROR=ON` by default and let users disable it locally. I think >> that would encourage developers to care more about warnings? Release >> branches could have it disabled somehow, to accommodate for the changing >> compiler environments. Was this discussed before? > > I don't remember whether this was discussed before. I think it would be a > good topic for llvm-dev. > > It would be nice to have at least one buildbot using -Werror, with some > specific blessed compiler version, but for that to not get annoying I think > we'd need to surface those warnings in code reviews, and even if we have > system a for that today it doesn't seem to be working very well. Not sure we'd be much worse off than we are today if some buildbots enabled -Werror without phabricator pre-commit integration. I'm pretty sure we've had -Werror buildbots in the past in that situation. In some ways it'd be ideal if it were a bootstrap build so we ensure LLVM's always warning-clean with current Clang, but that means longer turnaround. (wonder if we could configure buildbots to fail on warnings, but not enable -Werror - so the buildbot would also still produce other useful results, rather than halting on the first warning & potentially hiding/delaying other useful findings) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D116313/new/ https://reviews.llvm.org/D116313 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
