| 1 Mar 2024 |
Marc Jakobi | Download Vollmacht_fuer_Paketabholung.pdf | 19:10:38 |
@jade_:matrix.org | In reply to @johannes.kirschbauer:scs.ems.host just having fun today: using "nix#nix-clangStdenv" i get SIGSEGV (Address boundary error). But using "nix#nix" is fine.. no segfault. I just dont understand whats the difference in this convoluted flake.nix file in the cpp nix repository. Apart from that i would expect either both or none of the outputs to have segfault. gcc vs clang build | 19:17:24 |
@jade_:matrix.org | segv doesn't tell you anything, please coredumpctl debug if this is Linux and give the back trace (type bt into gdb) | 19:18:08 |
@jade_:matrix.org | thank you | 19:18:11 |
Alyssa Ross | C++ is not ABI-compatible between GCC and Clang, so you can't build some things with GCC, some with Clang, and then link them together. | 19:19:04 |
Alyssa Ross | If you override Nix to build with Clang, boost will still be built with GCC | 19:19:14 |
Alyssa Ross | So it's not expected to work. | 19:19:29 |
Alyssa Ross | If the Nix flake is doing this, it should stop. | 19:20:17 |
Alyssa Ross | ah, but it looks like it actually overrides stdenv in an overlay, so I guess that should work? | 19:21:10 |
Alyssa Ross | It's very hard to read | 19:21:12 |
Alyssa Ross | Ah, no, the overlay only applies the custom stdenv to Nix | 19:22:26 |
Alyssa Ross | So it won't work | 19:22:28 |
Alyssa Ross | cc Ericson2314 | 19:22:30 |
Alyssa Ross | (It might not actually be ABI compatibility that's the problem — I just remember that it's definitely not expected for this to work) | 19:31:41 |
John Ericson | Alyssa Ross: if you are using --gcc-toolchain as we are on Linux, isn't it supposed to be ABI compatible? | 21:27:53 |
John Ericson | (more precisely, it is using at least parts of libstdc++ and friends so the ABI is the same?) | 21:28:49 |
John Ericson | I don't fully get it but I always do clang development builds (mainly out of habit) and don't think I had a problem yet | 21:29:28 |