[PATCH] D54378: Add Hurd triplet

[Samuel Thibault via Phabricator via llvm-commits]

sthibaul added a comment.

For sure I consider that not breaking anything on other archs is a must :)

> accepting arbitrary suffixes is a mistake

Well, I was actually only talking about what the usual autoconf-provided config.guess script returns: i686-unknown-gnu0.9. The real triplet that we use in cross-compilation toolchains etc. is i386-pc-gnu or i386-unknown-gnu.

> If you absolutely must use this triple

Mapping i386-pc-gnu to i386-pc-hurd-gnu internally is fine for me. Actually strictly speaking it should be i386-pc-gnu-gnu, because the system is really called GNU. That's what we use inside glibc for instance. Would that be enough to keep the current strategy, or should it be still kept inside the clang driver? I'm asking mostly because since I know next to nothing on the llvm architecture, I don't really know how to do that other approach.

>   Explanation of all variations a Hurd triple can take (additional flags, variants from other compilers)

I don't think there is currently any  variant beyond i[3456]86-{pc,unknown}-gnu[0-9.]*

>   What does the OS/Env "gnu" means (does it mandate glibc, can it be used with some other libraries, etc)

Currently it mandates glibc, yes, there is no plan to make the GNU system use another library than the GNU library :)

> And we'd also have to understand the actual conflict of the OS/Env "gnu" alternatives.

I at least know about {linux,window,kfreebsd}-gnu{,eabi,eabihf,-elf}


https://reviews.llvm.org/D54378