Nix Cross Compiling | 578 Members | |
| 129 Servers |
| Sender | Message | Time |
|---|---|---|
| 7 Oct 2021 | ||
| plain clangStdenv with ld.bfd is fine more or less since it works fine for C | 16:34:05 | |
| 8 Oct 2021 | ||
| (xposting from main) | 00:48:22 | |
| does anyone uh know how i can tell libstdc++ to compile with more debug symbols? Id like to figure out if it will be easier to figure out what I want to figure out by looking at stuff with gdb or trying to read the source... doing that by itself might not be that hard (if i could find where in the guts of gcc it is...but i wouldnt know how to do to do the stdenv override either | 00:48:23 | |
libstdc++ is a part of gcc package. as opposed to glibc it does not have separateDebugInfo = true;. Perhaps it should? | 06:34:20 | |
| note that libstdc++ contains a small bit of support around header-mostly STL. maybe it will be enough to add debugging symbols into it's user? (library or final executable) | 06:35:43 | |
| trofi: so i should recompile gcc with debug symbols? (or lower optimization?) | 16:47:03 | |
| specifically my problem is that i want to inspect the internals of a basic_streambuf object in gdb | 16:47:33 | |
| and gdb says <incomplete type> | 16:47:38 | |
| Does Nix support Canadian cross-compilation? | 17:00:32 | |
| how's that different to the normal case? | 17:20:04 | |
| pie_: don't know yet, sorry. But plan to find out for sure in following weeks :) | 17:59:00 | |
| 9 Oct 2021 | ||
gdb's info sharedlibary always screws me upthe star in the read column means NOT read | 00:02:23 | |
https://gcc-help.gcc.gnu.narkive.com/LHU6EUxd/compiling-only-libstdc-v3-with-debugging-symbols lead me to https://gcc.gnu.org/onlinedocs/libstdc++/manual/configure.html , which suggests adding --enable-libstdcxx-debug to the configureFlags for gcc might get me what i want | 00:32:21 | |
| that manual page also gives this suggestion:
| 00:32:49 | |
| I'm not quite sure that implies what I'm saying; but it would be nice if we could build libstdc++ in a separate derivation. (somehow noting the tight coupling of a c++ stdlib and its compiler | 00:33:35 | |
| * I'm not quite sure that implies what I'm saying; but it would be nice if we could build libstdc++ in a separate derivation. (somehow noting the tight coupling of a c++ stdlib and its compiler) | 00:33:37 | |
In reply to @siraben:matrix.orgit's doable, but there's no ready made solution afaik | 09:32:38 | |
| John Ericson talked about this | 09:34:08 | |
| /nix/store/9lz65rckh7zhg4j76qzmiy0vr9hb0wa1-gcc-10.3.0/lib/debug/libstdc++.so.6 gets created but readelf still shows no debug sections on it... | 12:29:07 | |
| Anyone see anything wrong with this?
| 12:34:21 | |
| siraben if you call stage.nix yourself it should work | 16:46:25 | |
| Can I disable separated outputs on gcc? | 21:32:35 | |
| (I think) When I try to add keepTreeHook, I run into:
| 21:33:27 | |
| 11 Oct 2021 | ||
| We really need some mechanism to disable optional dependencies on systemd for musl | 09:53:35 | |
| or I suppose maybe we could try OpenEmbedded's patches | 10:12:30 | |
| not a friendly response last time that was tried though D: https://github.com/NixOS/nixpkgs/pull/37715 | 10:17:58 | |
| I saw a pr that introduced somthing to replace udev | 10:42:51 | |
| If we reverted to actually use udev as an input instead of systemd when it makes sens, we could overlay udev with this replacement when on musl | 10:44:04 | |
| if you could find that PR that would be very interesting to me! | 10:51:29 | |
| I was just packaging libudev-zero, which looks like it'll solve my problem, but there's no PR for that so it can't have been what you saw | 10:51:59 | |