| 1 Dec 2025 |
Grimmauld (any/all) |  Download image.png | 10:12:53 |
Grimmauld (any/all) | and we are building | 10:12:55 |
Grimmauld (any/all) | ghpzin: stdenv built with those patches, time to build against the VM | 11:36:22 |
Grimmauld (any/all) | it does eval at least, but it'll take a few hours to have useful info | 11:38:23 |
Grimmauld (any/all) | man-db wants groff, and yes i get that break too | 12:24:32 |
Grimmauld (any/all) |  Download image.png | 12:48:35 |
Grimmauld (any/all) | patchutils = prev.patchutils.overrideAttrs (old: rec {
version = "0.4.4";
src = final.fetchFromGitHub {
owner = "twaugh";
repo = "patchutils";
tag = version;
hash = "sha256-+gcVfNcAdcHTndHdarilbjIVsDJH6NTQsoQ5NN7veHU=";
};
nativeBuildInputs = old.nativeBuildInputs or [ ] ++ [
final.gettext
final.gnulib
final.autoconf
final.automake
];
postPatch = ''
./bootstrap
'';
});
Not too bad, seems 0.4.4 just works (though i did do a github fetch and reconf, the url we normally use didn't work for some reason)
| 12:49:28 |
Grimmauld (any/all) | opensp> dcigettext.c: At top level:
opensp> dcigettext.c:154:7: error: conflicting types for ‘getcwd’; have ‘char *(void)’
opensp> 154 | char *getcwd ();
opensp> | ^~~~~~
opensp> In file included from dcigettext.c:67:
opensp> /nix/store/xi34jjsk1y2n6pfykghiar2hlh4s16h4-fortify-headers-1.1alpine3/include/unistd.h:65:27: note: previous definition of ‘getcwd’ with type ‘char *(char *, size_t)’ {aka ‘char *(char *, long unsigned int)’}
opensp> 65 | _FORTIFY_FN(getcwd) char *getcwd(char * _FORTIFY_POS0 __s, size_t __l)
opensp> | ^~~~~~
opensp> dcigettext.c: In function ‘libintl_dcigettext’:
opensp> dcigettext.c:544:17: error: too many arguments to function ‘getcwd’; expected 0, have 2
opensp> 544 | ret = getcwd (dirname, path_max);
opensp> | ^~~~~~ ~~~~~~~
opensp> dcigettext.c:154:7: note: declared here
opensp> 154 | char *getcwd ();
opensp> | ^~~~~~
opensp> make[2]: *** [Makefile:175: dcigettext.o] Error 1
opensp> make[2]: Leaving directory '/build/OpenSP-1.5.2/intl'
opensp> make[1]: *** [Makefile:384: all-recursive] Error 1
opensp> make[1]: Leaving directory '/build/OpenSP-1.5.2'
opensp> make: *** [Makefile:273: all] Error 2
opensp looks painful... Dead project since 2005, yet po4a depends on it.
| 12:59:33 |
ghpzin | opensp builds fine on non-musl.
patchutils would one more fix for older version, otherwise fetchpatch won't work. | 13:13:35 |
ghpzin | opensp builds fine on non-musl.
patchutils would need one more fix for older version, otherwise fetchpatch won't work. | 13:13:56 |
ghpzin | opensp builds fine on non-musl.
patchutils would need one more fix for older version, otherwise fetchpatch won't work. I actually stumbled on conflicts with fortify-headers a few times, looked at alpine sources and there is a r5 version instead of r3 we have. But updating it never fixed anything. | 13:19:03 |
| matthewcroughan changed their profile picture. | 14:57:47 |
Fabián Heredia | Good news, got a patch that gets db{48,53,60,62} building with gcc15: https://github.com/NixOS/nixpkgs/pull/466885
That was the last big breakage :3 | 17:30:49 |
Fabián Heredia |  Download image.png | 17:39:26 |
Fabián Heredia | Could someone with eval priviledges do a last eval on the gcc15 jobset? https://hydra.nixos.org/jobset/nixpkgs/pr-440456-gcc15
Should only be about 7k jobs that failed due to db53. | 17:39:38 |
K900 | Done | 17:40:00 |
Fabián Heredia | Thanks <3 | 17:40:08 |
Grimmauld (any/all) | current fallout with musl on gcc 15: | 18:05:31 |
Grimmauld (any/all) |  Download image.png | 18:05:33 |
Grimmauld (any/all) | i guess db is non-musl too and fixed | 18:05:56 |
ghpzin | patchutils, mbetls, liblapack, opensp, dbus and python313Packages.virtualenv should be musl only (unless they failed with something else)
vde2, gfxstream, unixODBCDrivers.mariadb and webrtc-audio-processing_1 affect both
webrtc-audio-processing_1 had PR with fix somewhere, for other 3 things I have fixes, will do PRs tomorrow. | 18:30:42 |
ghpzin | patchutils, mbedtls, liblapack, opensp, dbus and python313Packages.virtualenv should be musl only (unless they failed with something else)
vde2, gfxstream, unixODBCDrivers.mariadb and webrtc-audio-processing_1 affect both
webrtc-audio-processing_1 had PR with fix somewhere, for other 3 things I have fixes, will do PRs tomorrow. | 18:34:54 |
Grimmauld (any/all) |
- virtualenv: test fail (
tests/unit/test_file_limit.py::test_too_many_open_files - assert 'Too many open files')
- mbedtls: test fail (
psa_crypto_pake-suite (Failed))
- vde2: something about invalid pointer types with
ssize_fun
- maria odbc: missing ‘#include <cstdint>’ i think? fails on
error: ‘uint8_t’ does not name a type, should be trivial
- opensp: symbol collision with libiberty i think?
error: too many arguments to function ‘getcwd’; expected 0, have 2
- gfxstream: same as maria thingy, missing cstdint for
uint8_t
- audio processing: same again, this time for
unint32_t
- patchutils: needs a bump to 0.4.4, but we have patchutils 3 around still...
| 18:50:38 |
Grimmauld (any/all) | those are the obvious ones | 18:50:58 |
Grimmauld (any/all) | the missing includes should be trivial, but there may be more issues than just that | 18:51:52 |
hexa | too many open files on what system? | 19:00:35 |
hexa | wait, it tests for that?! | 19:00:54 |
hexa | patchutils are used in fetchpatch and fetchpatch2, at least two versions | 19:01:53 |
Fabián Heredia | As an informal poll I feel like the gcc15 change is ready for staging.
Would anyone oppose at the moment or is there something that should be addressed before merging into staging? (More fixes can be sent before and during the staging-next cycle) | 21:16:01 |
Grimmauld (any/all) | my reaction was the same, and it might make sense to just skip that... | 21:51:10 |