| Summary: | media-gfx/openscad-2025.05.01_pre fails compiliton caused by manifold | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Karl Ernst Brunk <ke.b> |
| Component: | Current packages | Assignee: | Paul Zander <negril.nx+gentoo> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | ke.b, proxy-maint |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
build.log
emerge--info actual use flags |
||
Created attachment 928835 [details]
emerge--info
Created attachment 928836 [details]
actual use flags
x86_64-pc-linux-gnu-g++: fatal error: Killed signal terminated program cc1plus compilation terminated. This would point to running out of memory? Check dmesg. Please re-open if you have checked dmesg for out of memory messages. Hi Paul, sorry - i was one day out. memory ? - i dont think so and can't see any message. Its also very unlikelely while other big stuff like rust ist doing no problem. BUT. I tried to downgrade all direct dependencies (manifold, cgal) corresponding to the timeline of the last successful run of the 2024-version. And that gets strange. So even with those downgrades the 2024-version is not compilling anymore. Digging deeper - i tried also downgrading boost (1.88->1.87) because it is involved. No change. So at the moment i have no clue. cheers Karl You are running -j24 with 32 GiB of ram. Add -flto to that and the fact that the error message is what you usually see with being out of memory, I still don't see anything indicating this not being a out of memory. Building manifold also peaks around 31 GiB memory consumption around with -j24 -flto=24 where you see the compiler being killed. So yeah, lower the parallel jobs, lower the lto jobs, add swap and see if that fixes it. Hi again, i am so sorry to waste your time and me having a lack of imagination. I really did some research when multi-threads came up and did my own configuration with every new system. So 32Gb and 24 Threads was even with compiling everything on a RAM-disk never a problem - till this case came up. Even after rust forced me to give 13Gb of space. Now openscad and you teached me my lesson. And it was so close. When i give up the RAM-drive and lower the treads to !! 23 !! its compiling fine. With RAM-drive i have to set -j18 . I still wonder how it did run fine 2024. Maybe i forgot to mount the RAM-disk. So you where right and i have to apologize. Maybe it whould be a good idea to limit the threads through the ebuild or give a warning. sorry Karl |
Created attachment 928834 [details] build.log the new openscad version - but also the previous one - fail with the new version of manifold-3.02_pre20250330