| 15 Nov 2024 |
emily | but there were complications with it needing to be / for something something something | 12:03:47 |
emily | I forget, it was a few weeks ago | 12:03:51 |
emily | if you mean -I/usr/include, uhh | 12:03:57 |
emily | well (a) that's your fault for passing that and (b) we have a joke of a cc command-line parser in the wrappers to try and filter those out | 12:04:14 |
emily | which ALSO broke Darwin and I had to change it | 12:04:21 |
emily | (because on Darwin it's relative to the SDKROOT, so e.g. -L/usr/lib/swift is benign) | 12:04:38 |
p14 | I was under the impression the compiler driver searched usr lib of its own volition | 12:04:41 |
emily | it searches <sysroot>/lib | 12:04:53 |
p14 | https://github.com/llvm/llvm-project/blob/d7b669b3a30345cfcdb2fde2af6f48aa4b94845d/clang/lib/Driver/ToolChains/Gnu.cpp#L2930
Oh, right. | 12:05:11 |
p14 | And, we're not passing sysroot then I take it. | 12:05:22 |
emily | like I said apparently it has to be / but I forget why | 12:05:36 |
emily | (on Linux) | 12:05:39 |
p14 | that sounds odd, what's the point in having a sysroot parameter if that was true? | 12:05:53 |
emily | it was an arcane reason. ask trofi :) | 12:06:07 |
emily | In reply to @emilazy:matrix.org (because on Darwin it's relative to the SDKROOT, so e.g. -L/usr/lib/swift is benign) one possible path I've considered is making this behaviour configurable for all platforms. we pass -treat-all-those-user-specified-paths-as-sysroot-relative-even-though-this-isnt-darwin -sysroot /directory/tree/with/empty/stub/include/and/libs/directories in the wrapper | 12:06:18 |
emily | then we're safe from /usr/lib and don't need the broken purity hack CLI parsing in the wrapper | 12:06:32 |
emily | might work. dunno. | 12:06:39 |
emily | anyway, I think passing -nostdlibinc for non-Darwin targets in the wrapper is the conservative option for now, but we should definitely flag up to the relevant people that it no longer appears necessary for libc++ | 12:07:50 |
p14 | Bah, pkgsStatic.pkgsLLVM.hello-cpp.stdenv.cc/bin/x86_64-unknown-linux-musl-clang doesn't work, because it fails to find ld. -ld-path=x86_64-unknown-linux-musl-ld fixes it. | 12:09:09 |
p14 | * Bah, pkgsStatic.pkgsLLVM.hello-cpp.stdenv.cc/bin/x86_64-unknown-linux-musl-clang doesn't work, because it fails to find ld. --ld-path=x86_64-unknown-linux-musl-ld fixes it. | 12:09:17 |
p14 | "Bah", because, I expected it to already look for the triple-prefixed ld... | 12:09:31 |
emily | pkgsStatic.pkgsLLVM even does something sensible? | 12:09:36 |
emily | In reply to @p14:matrix.org "Bah", because, I expected it to already look for the triple-prefixed ld... wanna revive the "prefix all compilers" PRs? 😆 | 12:09:45 |
p14 | Looks like it does, sort of? :) | 12:09:48 |
emily | In reply to @trofi:matrix.org At least on linux the attempt to use non-standard sysroot broke DT_RUNPATH lookups for libraries not explicitly mentioned in the linking process: https://github.com/NixOS/nixpkgs/pull/213185. Transitive libraries are searched in sysroot only and things break. p14: here's a pointer to the previous discussion | 12:10:44 |
emily | with the conclusion of basically "we should add more knobs upstream and then twiddle only the ones we need" | 12:11:08 |
emily | which would give us a path to dropping -nostdlibinc without regressing things | 12:11:16 |
p14 | This can of worms is too big to eat. I'm going to leave nostdlibinc in, and send the PR to push it into the wrapper at least. | 12:13:42 |
p14 | https://github.com/NixOS/nixpkgs/pull/356162 | 12:22:14 |
p14 | In reply to @p14:matrix.org
I've come up with a possible use case for this abandoned PR https://github.com/NixOS/nixpkgs/pull/355000
Being able to put link-only flags in it for clang: warning: argument unused during compilation: '-pie' [-Wunused-command-line-argument] would be useful. Sensible?
An alternative route would be to bracket these flags in --start-no-unused-arguments and --end-no-unused-arguments.
Any opinions on this one? Introduce cc-cflags-link, and then wire link-specific flags into it? Or ignore unused args in more places? | 12:24:13 |