compnerd wrote: > > I don't think we should condone anything that assumes fetching WinSDK/MSVC > > derived files from a third party provider or github repo or anything like > > that. > > I agree, but can we condone, or even redistribute, a script that > post-processes the files this way, so anyone can replicate this setup for > their personal use? > > Looking at the linked github repo (which as you say may be violating a > license), it really makes me question the entire Unix /usr/(include|lib) > design philosophy, of having one single, global header and library search > directory. But, that is what people do in practice. :) > > > Perhaps one thing we could do is maybe relax that assumption, and see if we > > can shift some of the burden from users to the driver by having some form > > of case-insensitive lookup when we are targeting `x86_64-pc-windows-msvc`? > > I recall @compnerd asked for this at some point in the past, but perhaps he's > just adopted the `ciopfs` mount solution. Anyway, [my search didn't find an > existing > issue](https://github.com/llvm/llvm-project/issues?q=is%3Aissue%20state%3Aopen%20case%20insensitive), > so I'll file a new one for that. ... Filed as issue #161400 .
I ended up pursing a hybrid approach. I ended up with modifications to the compiler to support that partially since it was needed for the (clang) modules support. However, I also used a forest of symlinks to get as much as possible done more easily. I haven't had to support that in a while and I think that at this point, my approach would be to use a VFS overlay rather than the symlinks. The linker would need the symlinks for the import libraries still though. https://github.com/llvm/llvm-project/pull/96417 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
