Summary: | dev-python/visual-5.03_rc1 does not compile with dev-libs/boost-1.37.0-r1 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Sebastian Luther (few) <SebastianLuther> |
Component: | Current packages | Assignee: | Tiziano Müller (RETIRED) <dev-zero> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | cpp+disabled, dev-zero, djc, grozin, joost.ruis, python |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | All | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 181653 | ||
Attachments: |
build.log
patch that fixes dev-libs/boost header https://svn.boost.org/trac/boost/changeset/53731/sandbox-branches/bhy/py3k modified to apply cleanly against boost-1.39.0 always use correct boost version always use the correct boost version |
Description
Sebastian Luther (few)
2009-05-24 05:35:52 UTC
Created attachment 192262 [details]
build.log
That's not emerge's build.log but the log from work/../src/
(In reply to comment #0) > /usr/include/boost/python/detail/translate_exception.hpp:56: Fehler: »e« > wurde in diesem Gültigkeitsbereich nicht definiert Je ne le comprends pas. Utilisez LC_ALL=C, s'il vous plaît. (In reply to comment #2) > (In reply to comment #0) > > /usr/include/boost/python/detail/translate_exception.hpp:56: Fehler: »e« > > wurde in diesem Gültigkeitsbereich nicht definiert > > Je ne le comprends pas. Utilisez LC_ALL=C, s'il vous plaît. > Sorry for that, it means: "e" was not declared in this scope. OT: The ebuild component is for new ebuilds, not for bugs involving ebuilds. Created attachment 199811 [details, diff]
patch that fixes dev-libs/boost header
This patch for dev-libs/boost should fix it.
This looks like it needs boost patched. Adding maintainers. Same here: dev-python/visual-5.12 fails to compile with dev-libs/boost-1.39.0. Error messages are the same (only in another language :-) This is Boost bug https://svn.boost.org/trac/boost/ticket/2582. An easy workaround: boost is slotted. I installed and eselected boost-1.35, after that visual-5.12 emerged without problems. (In reply to comment #8) > An easy workaround: boost is slotted. I installed and eselected boost-1.35, > after that visual-5.12 emerged without problems. > Yes, it merges, but executing examples in /usr/share/doc/visual-5.12/examples/ results in segmentation faults for me. The same happened to me with visual-5.13, recently added to portage. (In reply to comment #9) > (In reply to comment #8) > > An easy workaround: boost is slotted. I installed and eselected boost-1.35, > > after that visual-5.12 emerged without problems. > > > > Yes, it merges, but executing examples in /usr/share/doc/visual-5.12/examples/ > results in segmentation faults for me. Same here. Apparently this is a boost/python problem (it's reproducible with all versions of dev-python/visual). I will attach a patch for boost which fixed the problem for me. > > The same happened to me with visual-5.13, recently added to portage. > Created attachment 210866 [details, diff] https://svn.boost.org/trac/boost/changeset/53731/sandbox-branches/bhy/py3k modified to apply cleanly against boost-1.39.0 I added the patches to the 1.39 and 1.40 ebuilds in my overlay [1]. 1.41 has both included. [1] git://github.com/few/few-s-gentoo-overlay.git dev-libs/boost maintainers: Please apply these patches. ping boost maintainers I've emerged boost-1.41 which just appeared in the tree, then re-emerged visual-5.13, and it works. Very nice (though I don't use it often). So, we can depend on >=dev-libs/boost-1.41, buth this does not solve the problem - boost is slotted, and a user can eselect some older boost. Is it possible for an ebuild (visual in our case) to use a specific slot of boost, ignoring what a user eselects? Something similar is done in the case of wxwidgets (eselect settings ignored), through wxwidgets.eclass. But, of course, it would be even better to fix all versions of boost present in the tree. (In reply to comment #15) > I've emerged boost-1.41 which just appeared in the tree, then re-emerged > visual-5.13, and it works. Very nice (though I don't use it often). So, we can > depend on >=dev-libs/boost-1.41, buth this does not solve the problem - boost > is slotted, and a user can eselect some older boost. Is it possible for an > ebuild (visual in our case) to use a specific slot of boost, ignoring what a > user eselects? Something similar is done in the case of wxwidgets (eselect > settings ignored), through wxwidgets.eclass. But, of course, it would be even > better to fix all versions of boost present in the tree. > I'll write you a patch that makes it ignore the eselect setting (which is recommended way for all ebuilds) and I hope we get the other boost versions fixed too... Created attachment 211902 [details, diff] always use correct boost version This didn't workout very well. The configure script doesn't provide a way to specify the include and library directory for boost (it doesn't even check if boost is present). You should really make upstream fix that. As long this isn't fixed, I can only offer this hack. Once it's fixed you should use something like the patch on bug 295474. Created attachment 211906 [details, diff]
always use the correct boost version
Thank you very much. In cvs. |