| Summary: | media-sound/ardour-4.6: sed fails. configure also fails when Python3 default | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | wjn |
| Component: | Current packages | Assignee: | Andreas Schürch <nativemad> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | gmt, proaudio |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: | various_ardour-4.6.ebuild_fixes.patch | ||
|
Description
wjn
2016-01-13 11:43:55 UTC
@ proaudio project: please add "|| die" to each sed command. This is qa relevant. Thanks for reporting wjn! 1) It looks like that $OPTFLAGS variable got some nice cultural influence from Japan! :-) Have you set something in CPU_FLAGS_X86 in make.conf? Do you get better results if you just set CPU_FLAGS_X86="sse"? Can you please also post the output of "emerge --info"? ...I guess it will be enough to just set OPTFLAGS="" at the start of src_prepare(). But to get sure please post the above output anyway... 2) I can't reproduce it atm, as I don't have python-3.5 around yet... But your proposal sounds good imho, as that would also work around Bug #566866 I will have a look at it. Thanks also for the input Lars! :-) (In reply to wjn from comment #0) > 1. sed fails Thanks for tracking this down! > 2. src_configure fails when Python 3.5 is default To be clear, this occurs when python35 is eselect-python-ed. This ebuild doesn't get the magic PYTHON_TARGETS USE_EXPAND behavior that otherp python-any-r1 ebuilds do because the magic happens in python-any-r1_pkg_setup but the ebuild implement that to and never calls the overridden inherited* implementation. So, even if it's not strictly orthogonal to issue #1, I'd strongly opine that the following patch should be applied in addition to the fix described by OP: --- media-sound/ardour/ardour-4.6.ebuild.orig 2016-01-13 12:13:45.833062068 -0800 +++ media-sound/ardour/ardour-4.6.ebuild 2016-01-13 12:12:29.333885416 -0800 @@ -73,6 +73,7 @@ pkg_setup() { if has_version \>=dev-libs/libsigc++-2.6 ; then append-cxxflags -std=c++11 fi + python-any-r1_pkg_setup } src_prepare(){ -------------------------- * fairly pointless TLDR terminologico-anal digression: OOP terminology like "overridden" arguably do not apply to eclasses because they are not classes. But, they are called eclasses and the word "inherit" is used to establish the relationship I am describing... so what is the right word for the implementation we blew away with this function? IMO it's fine, so long as we all remember we are speaking metaphorically, and that the key to not overextending this particular metaphor is to never ask: "if these are the 'eclasses,' where are the 'eobjects?'" Created attachment 422836 [details, diff] various_ardour-4.6.ebuild_fixes.patch (In reply to Andreas Schürch from comment #2) > Thanks for reporting wjn! > 1) > It looks like that $OPTFLAGS variable got some nice cultural influence from > Japan! :-) I don't think that's it. Unless I got some sort of anime virus :) The actual problem reported is unterminated sed command @ character 41, for me. My OPTFLAGS="sse mmx" at that point in the ebuild (I added an ewarn to be sure) so thats... V 0123456789.11.14.17.20.23.26.29.32.35.38.41 s/flag_line\ =\ o.*/flag_line\ =\ \": sse mmx just some place holders\" uhh... *blink*... *ponder* Ooooh, haha. We are just paying the standard bash word-splitting sanity-tax on this one. Once we dequote we never re-quote so "mmx" is the beginning of a new argument to sed.... so here's a patch with that fixed, and my patch, plus everyone else's ideas (they all LGTM), put together. Ok, I've seen the things now and committed the fixes in the 4.6 and 9999 ebuilds. Thanks all! |
