yaxunl added a comment. In D111443#3062139 <https://reviews.llvm.org/D111443#3062139>, @eugenis wrote:
> Right, but a cache for SanitizerArgs is not enough to avoid repeated > diagnostics, is it? Ex. if I request a non-existing sanitizer, I think I > would get errors from host arg parsing, as well as from each of device1 and > device2, because each device will have a unique ArgList. > > Is it even possible for the driver to introduce new diagnostics in offload > SanitizerArgs parsing? Is it possible to catch those cases ahead of time, > when parsing SanitizerArgs for the first time, by looking at a target triple > or something? That would be the most user friendly approach. I think it makes sense to assume offloading toolchain will not introduce extra diagnostics inside SanitizerArgs than the default toolchain, therefore it should be OK to let the default toolchain check sanitizer args for once. I will try make that change. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D111443/new/ https://reviews.llvm.org/D111443 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits