=== techalchemy_ is now known as techalchemy | ||
RikMills | doko: do you have an idea what might need to be done to resolve this? https://paste.ubuntu.com/p/S7Q5y3sYb5/ | 11:12 |
---|---|---|
doko | RikMills: the llvm-12 build on riscv64 just finished 2h ago. maybe just retry? | 11:14 |
RikMills | doko: that is the 1st thing I checked in the build log. it used that llvm build :/ | 11:16 |
RikMills | In fact that is why I did that build. to test if the new llvm-12 solved the issue | 11:18 |
doko | we were using gcc-11 on riscv64 to build llvm-12. we could revert that now. the needed patch is backported to gcc-10 | 11:18 |
RikMills | shal I report a bug? | 11:19 |
RikMills | *shall | 11:19 |
doko | I don't understand why libstdc++6 doesn't satisfy that symbol ... | 11:20 |
doko | LocutusOfBorg: ^^^ | 11:20 |
ricotz | hi :), could you retry the autopkgtest for vala here? https://people.canonical.com/~ubuntu-archive/proposed-migration/update_excuses.html#vala | 11:21 |
RikMills | doko: only ref I found online was https://www.mail-archive.com/gcc-patches@gcc.gnu.org/msg255111.html | 11:21 |
RikMills | but I am slightly out on my depth on that, so not sure if relevant | 11:22 |
doko | these baslines only affect libstdc++ tests | 11:23 |
ricotz | ah sorry, it was retried already :) | 11:23 |
doko | RikMills: libllvm12_12.0.0~++rc3-3_riscv64.deb has the correct versioned dep: libstdc++6 (>= 11). odd | 11:28 |
RikMills | doko: if it is relevant, I had a successful build about 24hrs before my 1st fail, and only significant difference I can see in the gcc/llvm stack on that one was libstdc++6:riscv64 (11-20210313-0ubuntu1) used on the successful build vs libstdc++6:riscv64 (11-20210317-0ubuntu1) | 11:29 |
RikMills | on the fail | 11:29 |
RikMills | https://paste.ubuntu.com/p/5WV9MW8Z4Q/ | 11:31 |
RikMills | ^ diff of the deps installed for the build | 11:31 |
doko | RikMills: indeed, _ZNSt9once_flag11_M_activateEv is dropped in -17 | 11:39 |
doko | RikMills: ahh, I remember, that's PR 99341. libstdc++: Remove symbols for new std::call_once implementation. so the action to wait for today's gcc-11 build, then rebuild llvm-toolchain-12. I don't want to back-out that patch | 11:45 |
doko | so will take a few days with these multi-day build times | 11:46 |
RikMills | doko: and the new llvm-12 build build with -13 on riscv64 as far as I can see. probably because -17 was taking huge time to finish | 11:46 |
RikMills | doko: that is great, and makes sense. thank you :) | 11:47 |
doko | I'll upload a doxygen build with 11 to fix the doxygen issue for now | 11:48 |
RikMills | thanks! | 11:49 |
=== sem2peie- is now known as sem2peie | ||
LocutusOfBorg | doko, I plan a new llvm-toolchain-12 upload in Debian | 15:09 |
LocutusOfBorg | so I'll just wait for gcc-11/riscv64 to finish and then sync? | 15:09 |
doko | LocutusOfBorg: please revert back to b-d gcc-10 (>= 10.2.1-21) | 15:21 |
LocutusOfBorg | doko, for debian too? | 16:14 |
LocutusOfBorg | its in sid now... | 16:14 |
LocutusOfBorg | you mean drop this line g++-11 [riscv64], | 16:15 |
doko | LocutusOfBorg: does it matter? unstable doesn't have gcc-11 either | 16:15 |
LocutusOfBorg | yes so just drop that line? | 16:15 |
doko | sur | 16:15 |
doko | well, and don't use 11 in the rules | 16:17 |
LocutusOfBorg | yes I reverted that already | 16:25 |
LocutusOfBorg | thanks | 16:25 |
LocutusOfBorg | uploaded in sid, will sync later, we don't need to wait for gcc-11 anymore I guess | 16:29 |
=== halvors1 is now known as halvors |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!