Summary: | dev-lisp/ecls-9* and ecls-10.2 fail to build on ppc | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | François Bissey <frp.bissey> |
Component: | New packages | Assignee: | Common Lisp Bugs <common-lisp> |
Status: | RESOLVED TEST-REQUEST | ||
Severity: | normal | CC: | frp.bissey, ppc |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | PPC | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
build log of ecls with gcc-4.3.4
build log of ecls with gcc-4.4.3 patch to ecls-10.4.1 ebuild |
Description
François Bissey
2010-03-10 22:57:58 UTC
Created attachment 223089 [details]
build log of ecls with gcc-4.3.4
Created attachment 223091 [details]
build log of ecls with gcc-4.4.3
Adding -std=c99 to my CFLAGS brings the building of ecls with gcc-4.4.3 at the same point than with gcc-4.3.4. Why you need this on ppc and not on x86 is beyond me. The next problem is why doesn't it bootstrap and why do I have all those warning about implicit declarations that don't appear on x86. I should make an update on that. Removing all my cflags except for -std=c99 resulted in a successful compilation with gcc-4.4.3. I haven't investigated fully which flag or combination of flags is guilty. Well, does it work now? (In reply to comment #5) > Well, does it work now? > I finally got to do a test of it today with maxima. It seems to work. So in short one of my CFLAGS (which aren't really ricers flags) is causing problem. I cannot point the finger on which one yet - this old G4 is updating some other packages at the moment and it will take a while. For some reason on ppc with gcc-4.4.3 -std=c99 is needed. OK I cornered it. ecls is allergic to altivec instructions. Appending "-mno-altivec -mabi=no-altivec" to the CFLAGS solve the problem. I will probably up an ebuild with the fix in the science overlay later on. Created attachment 230377 [details, diff]
patch to ecls-10.4.1 ebuild
So I ended up just patching the ebuild to add the appropriate flags.
I had a quick look at ecls sources but it would be quite difficult
to solve the problem in the source as far as I can see.
Upstream may or may not be interested in this problem.
@ppc team, can somebody approve the patch? I know nothing about the altivec instruction set. Thank you in advance. Can you try with ecls-16.0.0? It looks to be keyworded on ppc Dear me, talk about a zombie. I don't have a Gentoo ppc machine anymore I am afraid but I believe it would work out of the box on ppc64 (gentoo or not). |