| Summary: | sci-libs/blas-atlas-3.7.24 build fails on alpha, hppa: unrecognized command line option "-m32" to gcc | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Stefaan De Roeck (RETIRED) <stefaan> |
| Component: | New packages | Assignee: | Gentoo Science Related Packages <sci> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | armin76, jer |
| Priority: | High | ||
| Version: | 2006.1 | ||
| Hardware: | Alpha | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | |||
| Bug Blocks: | 178413, 180713, 189738 | ||
| Attachments: |
Output of "ebuild ... compile"
"emerge blas-atlas" for patched version |
||
|
Description
Stefaan De Roeck (RETIRED)
2006-12-24 00:40:55 UTC
Created attachment 104672 [details]
Output of "ebuild ... compile"
(In reply to comment #1) > Created an attachment (id=104672) [edit] > Output of "ebuild ... compile" > Hi Stefaan, Thanks for pointing this out. I'll see that I can get rid of the -m32 in atlas' build system for alpha. Hopefully that'll do. Best, Markus I've disabled the explicit 32/64 bit arch selector in the ebuild for alpha. However, atlas' build system will still do some auto-detection but will hopefully refrain from passing -m32 on alpha. Unfortunately, I can't test this myself. Hence, could you please test and let me know if it works for you! Thanks, Markus Created attachment 105275 [details]
"emerge blas-atlas" for patched version
Tried "emerge blas-atlas" after cvs up, unfortunately, there seems to be more to it :( It now fails on "-m64", and it seems to do so the first time in work/ATLAS/gentoo-build (during unpack), and persists on doing so during the compilation process.
Hi Stefaan, Thanks for the quick response! Somehow I knew that this would be happening since upstream has significantly overhauled their build system and "officially" only supports x86, x86-64 and sparc, AFAIK. I'll have a look at it and hopefully we can get rid of that pesky -mXX flag. Best, Markus Any news on this one? i got the same error but -m64 on 3.7.32 (In reply to comment #6) I assume you mean the same error as I had in comment #4 then? Unfortunately, no further news :( I think this one will have to wait for upstream, since the build system clearly has trouble detecting alpha properly. Even if I'd be able to get rid of the -m64 somehow there's still no guarantee that the build does anything sensible and may very well result in a pretty crummy library. We just discussed this among the sci devs and have tentatively decided to make the blas|lapack|cblas-reference ebuilds the default for the blas|lapack|cblas-virtual instead of the *-atlas ebuilds as presently. The reference ebuilds should be a much easier and faster compile and provide more than adequate libraries. Currently only the alpha keyword is missing for blas-reference lapack-reference cblas-reference Hence, if you guys could have a look and possibly keyword these that would be great. Otherwise, I'll file a separate bug about this when I have some more time on the weekend. Thanks, Markus Okay, i've keyworded those ebuilds ~alpha. Well, i've removed -m64 from blas-atlas and lapack-atlas, and everything seems fine. (In reply to comment #10) > Well, i've removed -m64 from blas-atlas and lapack-atlas, and everything seems > fine. > Thanks, let's close this one then. |