Summary: | [science-overlay] sci-mathematics/axiom-201005 fails to configure subpackage gmp | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Juergen Rose <rose> |
Component: | Current packages | Assignee: | Gentoo Science Mathematics related packages <sci-mathematics> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | dilfridge, tomka, xmw |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | URL to bezipped build.log |
Description
Juergen Rose
2010-09-11 08:48:07 UTC
build.log!!! Created attachment 248022 [details]
URL to bezipped build.log
Hi, I think the actual failure is the following: checking for working makeinfo... found checking whether to enable maintainer-specific portions of Makefiles... no configure: error: ABI=amd64 is not among the following valid choices: standard # # Subconfigure of GMP done # ------------------------ The subconfigure of gmp dies, but actually I never want to get it there, but use the systems gmp instead. Did you have dev-libs/gmp installed on that system at the time of emerging axiom? If not then it might simply be a missing dependency, otherwise I don't really know. Axiom bundles a lot of messy and severely broken software (like gnu common lisp) which will be hard to unbundle. Please try if it works if you have dev-libs/gmp installed. Thanks dev-libs/gmp was already ever installed: root@lynx:/root(2)# genlop -t dev-libs/gmp | tail Fri May 14 19:59:21 2010 >>> dev-libs/gmp-4.3.2 merge time: 2 minutes and 24 seconds. Mon Jul 5 07:47:45 2010 >>> dev-libs/gmp-5.0.1 merge time: 2 minutes and 32 seconds. Tue Sep 14 02:52:28 2010 >>> dev-libs/gmp-5.0.1 merge time: 2 minutes and 1 second. I've built this axiom on x86 and amd64 and cannot reproduce the failure. Let me know if you have any ideas. Fact: This configure of gmp that is called and fails in your build.log should not be called in the first place. The actual failure is a simple renaming of the content of the variable ABI which is properly fixed in the ebuild of gmp in the tree. I tried it on three amd64 boxes and got every the same error. Hallo Thomas, I reeemerged dev-libs/gmp, then I tried again to emerge axiom it fails with the same error. When I installed dev-libs/gmp-5.0.1, it was suggested to remove /usr/lib/libgmp.so.3, what I did. Could it be that you still have /usr/lib/libgmp.so.3 and that axiom-201005 need libgmp.so.3? Downgrading gmp is a no-go, see bug 338762. On request from QA I've masked the axiom-201005 in the overlay pending some better idea... Sorry guys. My bad. I adjusted the of gmp-dep of axiom last night because of this bug here. I also sent "RFC: Last rite axiom" to the science mailing list. Thank you for masking it. That is indeed the proper solution until we have figured out if upstream can fix it or we drop it. Juergen, after we had this discussion here I figured that you are right. This ebuild will not work with gmp-5, so I adjusted the dep, I did not think that there was a setup where an actual downgrade of gmp would take place, but more that you simply could not install axiom because of the missing dep. Sorry. Hi Thomas, forgiven. Juergen Hi, this is fixend in an ugly way though. I just bumped axiom to 201009 which, for now, has no keywords. It should build and run with gmp-5 installed. To do so it will build an internal copy of gmp-4 as a dependency of the interal copy of gnu common lisp. I see now chance to resolve this from the gentoo side, other than getting new unmasked gcl packages in the tree. Upstream considers it risky to depend on packages outside of their control (e.g. a system wide install of gmp). We are a bit stuck for now. Let's wait and see whether dev-lisp/gcl comes back to life. Closing. Thanks. Please follow bug 325067 for the future development of the axiom ebuild. I also removed the package.mask in the science overlay. |