| 23 Sep 2025 |
 hexa | 25.05 is until EOY | 16:29:09 |
| hexa | so slightly more than 3 months 😄 | 16:29:17 |
 aloisw | Right, the releases are usually at the end of the month. Maybe I'll move patches after cargoDeps in the next manual backport then, this way the following ones shouldn't conflict due to the line being out of context. | 16:30:47 |
 VladimÃr ÄŒunát | i.e. 25.05 is just past its half-life. | 16:30:57 |
 Grimmauld (any/all) | well, we didn't get another 24.11 staging cycle after 25.05 released, sooo the effective maintenance it gets might be less | 16:31:04 |
| hexa | yeah, staging cycle scheduling for three target branches is a mess | 16:32:25 |
| aloisw | It does not matter for Lix backports though. | 16:32:38 |
| VladimÃr ÄŒunát | The first missing merge is from May 18 | 16:32:41 |
| VladimÃr ÄŒunát | And the promised maintenance was until end of June? | 16:33:06 |
| VladimÃr ÄŒunát | Not great, I guess. | 16:33:13 |
| VladimÃr ÄŒunát | * The first missing merge is from May 18 (4a88dfe5c87) | 16:33:45 |
| aloisw | The aarch64-linux builder seems to be down, so x86_64-linux testing only for now. Should be enough though. | 16:34:22 |
| VladimÃr ÄŒunát | Most security fixes aren't huge rebuilds, fortunately. | 16:34:22 |
| hexa | the community builder? wfm | 16:34:39 |
| aloisw | No, the Lix builder. | 16:34:47 |
| hexa | ah | 16:34:54 |
| aloisw | I do not have access to the community builder. | 16:34:55 |
| hexa | https://nix-community.org/community-builders/ | 16:35:25 |
| hexa | just have to ask | 16:35:38 |
| aloisw | Well I don't expect Darwin-specific breakage from that PR, so it doesn't really matter. | 16:39:36 |
 Lun | Two more CMake 4 fixes https://github.com/NixOS/nixpkgs/pull/445579 | 17:03:51 |
| aloisw | Confirmed builds and tests work on x86_64-linux. Static build fails due to unrelated issues (nghttp3 trying to build a dynamic library and some weird test failure in nlohmann-json). | 17:59:53 |
| aloisw | nghttp3: https://github.com/NixOS/nixpkgs/pull/445602 | 18:07:46 |
| aloisw | Ah wait ngtcp2 has the same shit. | 18:08:12 |
| aloisw | ngtcp2-static-x86_64-unknown-linux-musl> /nix/store/yh6s1ihxlxcdb4ihxm7zvk9356jsl8nw-x86_64-unknown-linux-musl-binutils-2.44/bin/x86_64-unknown-linux-musl-ld: cannot find -lngtcp2_crypto_ossl: No such file or directory
ngtcp2-static-x86_64-unknown-linux-musl> /nix/store/yh6s1ihxlxcdb4ihxm7zvk9356jsl8nw-x86_64-unknown-linux-musl-binutils-2.44/bin/x86_64-unknown-linux-musl-ld: have you installed the static version of the ngtcp2_crypto_ossl library ?
ngtcp2-static-x86_64-unknown-linux-musl> /nix/store/yh6s1ihxlxcdb4ihxm7zvk9356jsl8nw-x86_64-unknown-linux-musl-binutils-2.44/bin/x86_64-unknown-linux-musl-ld: cannot find -lngtcp2: No such file or directory
ngtcp2-static-x86_64-unknown-linux-musl> /nix/store/yh6s1ihxlxcdb4ihxm7zvk9356jsl8nw-x86_64-unknown-linux-musl-binutils-2.44/bin/x86_64-unknown-linux-musl-ld: have you installed the static version of the ngtcp2 library ?
Yuck. | 18:11:06 |
| aloisw | Oh great. CMake will just shove the name you pass into target_link_libraries into the linker when there is no target with that name. So it works with dynamic library where the target is also called ngtcp2_crypto_ossl, but fails on static where it's called ngtcp2_crypto_ossl_static. | 18:17:33 |
| aloisw | Worked around by disabling the examples too. Static curl builds now. | 18:21:55 |
| aloisw | /build/source/tests/src/unit-locale-cpp.cpp:134:
TEST CASE: locale-dependent test (LC_NUMERIC=de_DE)
check if locale is properly set
/build/source/tests/src/unit-locale-cpp.cpp:142: ERROR: CHECK( std::string(buffer.data()) == "12,34" ) is NOT correct!
values: CHECK( 12.34 == 12,34 )
Is this supposed to work on musl? | 18:33:44 |
| aloisw |
LC_NUMERIC and LC_MONETARY also admit no variation by locale. Solving these problems requires a major overhaul, but the main missing prerequisite is involvement from users who want the functionality.
Seems like no, but I'm too tired to figure out how to disable that test in nlohmann_json. | 18:38:05 |
| aloisw | Actually they have a patch upstream (https://github.com/nlohmann/json/pull/4770), but I assume it's too late to squeeze a nlohmann rebuild in? | 18:45:41 |
