| Summary: | net-dns/avahi-0.6.10 fails to emerge | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Antti Aspinen <antti.aspinen> |
| Component: | New packages | Assignee: | Gentoo Linux bug wranglers <bug-wranglers> |
| Status: | VERIFIED WORKSFORME | ||
| Severity: | normal | ||
| Priority: | High | ||
| Version: | 2006.1 | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
Config.log
Config.log from avahi, this time without ffast-math. |
||
Changing the mistaken summary. Please, read the ebuild. If you wish to continue in this bug, then attach the config.log, the bug is useless without it. Plus make sure you've run python-updater after upgrading python. Created attachment 100940 [details]
Config.log
Uhm, here it is.
I noticed that emerging pygtk doesn't help it. Avahi still complains about it. Re-emerge python without -ffast-math and all the other junk in your C[XX]FLAGS and try again. (In reply to comment #5) > Re-emerge python without -ffast-math and all the other junk in your C[XX]FLAGS > and try again. > So, what you are saying is that I need to do freaking emerge -epv world because of this? I haven't never had any problems with --ffast-math. It's not junk. It's even in AMD recomended CFlags: "-O3 -march=athlon64 -ffast-math -funroll-all-loops -funit-at-a-time -fpeel-loops -ftracer -funswitch-loops -fomit-frame-pointer -pipe" Note that I have changed -O3 into -OS in my CFlag config. Tested if I am able have smaller as fast as O3 like binaries. Yeah yeah, I know, -O3 and -fomit-frame-pointer make debuggin impossible. Not that I don't need it but having funsafe-math-optimizations is great because it makes things work bit faster. I'll try dropping it out now, but I highly doupt it will change anything. (In reply to comment #6) > So, what you are saying is that I need to do freaking emerge -epv world because > of this? No, I told you to re-emerge python, not -epv world (though that'd be a bit more sane :P) > I haven't never had any problems with --ffast-math. It's not junk. > It's even in AMD recomended CFlags: "-O3 -march=athlon64 -ffast-math > -funroll-all-loops -funit-at-a-time -fpeel-loops -ftracer -funswitch-loops > -fomit-frame-pointer -pipe" Recommended by whom and for what purpose? o_O -ffast-math is a _stupid_ global flag, is not safe and breaks packages. man gcc <snip> -ffast-math Sets -fno-math-errno, -funsafe-math-optimizations, -fno-trapping-math, -ffinite-math-only, -fno-rounding-math and -fno-signaling-nans. This option causes the preprocessor macro "__FAST_MATH__" to be defined. This option should never be turned on by any -O option since it can result in incorrect output for programs which depend on an exact implementation of IEEE or ISO rules/specifications for math functions. </snip> > having funsafe-math-optimizations is great because it makes things work bit > faster. Faster and correct are two separate things... :P (In reply to comment #7) > (In reply to comment #6) > > So, what you are saying is that I need to do freaking emerge -epv world because > > of this? > > No, I told you to re-emerge python, not -epv world (though that'd be a bit more > sane :P) Thank god, I have done it already couple of times. Takes freaking three weeks to compile whole system again. > > I haven't never had any problems with --ffast-math. It's not junk. > > It's even in AMD recomended CFlags: "-O3 -march=athlon64 -ffast-math > > -funroll-all-loops -funit-at-a-time -fpeel-loops -ftracer -funswitch-loops > > -fomit-frame-pointer -pipe" > > Recommended by whom and for what purpose? o_O Recommended by AMD aka. Advanced Micro Devices: http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/32035.pdf I have used that cflag since I got Athlon64 like somewhat three years ago. > -ffast-math is a _stupid_ global flag, is not safe and breaks packages. > <snip> > This option should never be turned on by any -O option since it can result in > incorrect output for programs which depend on an exact implementation of IEEE > or ISO rules/specifications for math functions. > </snip> So? I have already read that junk when I first noticed that portage didn't like my Cflags. :) > > having funsafe-math-optimizations is great because it makes things work bit > > faster. > > Faster and correct are two separate things... :P > It seems so. :P Communism is fair but doesn't mostly work. Capitalism isn't fair but mostly works. ...Talking about python. I recompiled it and ran python-updater and tried to re-emerge avahi but it still doesn't get past that pygtk. I recompiled also pygtk which didn't have any effect. Created attachment 100949 [details]
Config.log from avahi, this time without ffast-math.
Config.log from avahi, this time without ffast-math.
Like I told ya, It doesn't have any effect in it. Problem is somewhere else I am afraid.
ffast-math haven't caused me problems, or at least I am not aware of those. I don't use Apache2, Compiling OO.o is masked on AMD64 anyway, I don't have postresql... GMPlayer works perfectly fine, ffmpeg works also. So I don't really get why people are having that much attitude problems towards ffast-math. Am I just so lucky not to have any problems? :P
(In reply to comment #8) > Recommended by AMD aka. Advanced Micro Devices: > http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/32035.pdf For what? Benchmarks? It makes no sense as global flag. > > This option should never be turned on by any -O option since it can result in > > incorrect output for programs which depend on an exact implementation of IEEE > > or ISO rules/specifications for math functions. > > </snip> > > So? I have already read that junk when I first noticed that portage didn't like > my Cflags. :) So? -ffast-math is NOT a safe global flag, it breaks things and we don't support it. It breaks python so that it fails when compiling OO.org, it breaks things that rely on *sane* math (as opposed to fast and broken math). > ...Talking about python. I recompiled it and ran python-updater and tried to > re-emerge avahi but it still doesn't get past that pygtk. I recompiled also > pygtk which didn't have any effect. What does python -c 'import gtk' produce? Does it compile if you run ssh localhost; emerge avahi? (In reply to comment #10) > (In reply to comment #8) > > Recommended by AMD aka. Advanced Micro Devices: > > http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/32035.pdf > > For what? Benchmarks? It makes no sense as global flag. Oh, well, I don't care about benchmarks. If AMD has a recomendation for Athlon64 I am more likely to belive them more than forums.gentoo.org. :D > So? -ffast-math is NOT a safe global flag, it breaks things and we don't > support it. It breaks python so that it fails when compiling OO.org, it breaks > things that rely on *sane* math (as opposed to fast and broken math). ok, ok, ok. I am aware of the problems with ffast-math already. > > ...Talking about python. I recompiled it and ran python-updater and tried to > > re-emerge avahi but it still doesn't get past that pygtk. I recompiled also > > pygtk which didn't have any effect. > > What does python -c 'import gtk' produce? Does it compile if you run ssh > localhost; emerge avahi? python -c 'import gtk': Traceback (most recent call last): File "<string>", line 1, in ? File "/usr/lib64/python2.4/site-packages/gtk-2.0/gtk/__init__.py", line 48, in ? from gtk import _gtk File "/usr/lib/python2.4/site-packages/cairo/__init__.py", line 1, in ? from _cairo import * ImportError: /usr/lib64/python2.4/site-packages/cairo/_cairo.so: undefined symbol: cairo_ps_surface_set_dpi_REPLACED_BY_cairo_surface_set_fallback_resolution ssh localhost; emerge avahi: "ssh: connect to host localhost port 22: Connection refused" And rest of what follows is avahi failing on pytgk. "checking for python module gtk... no configure: error: Could not find Python module gtk" Re-emerge cairo and pycairo, it's apparently broken. (In reply to comment #12) > Re-emerge cairo and pycairo, it's apparently broken. > Doing that made it downgrade glitz and cairo into some unknown version... I think my Beryl/GLX/Compiz went just broken... :D I am recompiling python and pygtk here too. Just got cairo and pycairo recompiled, it gives bit different error: python -c 'import gtk': Traceback (most recent call last): File "<string>", line 1, in ? File "/usr/lib64/python2.4/site-packages/gtk-2.0/gtk/__init__.py", line 48, in ? from gtk import _gtk ImportError: /usr/lib/libpangocairo-1.0.so.0: undefined symbol: cairo_scaled_font_get_font_options I guess, next I have to recompile pango... :) > I guess, next I have to recompile pango... :)
>
emerge pango made cairo and glitz into upgrade back to their current versions. Also it apparently fixed "python -c 'import gtk'" as it doesn't report anything back anymore.
I'll try emerging avahi now.
Okey, it passed that pygtk phase of config now. Thank you. :)
(In reply to comment #14) > > I guess, next I have to recompile pango... :) > > > > emerge pango made cairo and glitz into upgrade back to their current versions. > Also it apparently fixed "python -c 'import gtk'" as it doesn't report anything > back anymore. > > I'll try emerging avahi now. > Okey, it passed that pygtk phase of config now. Thank you. :) > Avahi compiled nicely. I'll close this bug. (In reply to comment #14) > emerge pango made cairo and glitz into upgrade back to their current versions. > Also it apparently fixed "python -c 'import gtk'" as it doesn't report anything > back anymore. Eh... If you are mixing ~arch and stable ebuilds, you should do it in such way that these upgrade/downgrade cycles don't happen, it breaks things. You apparently are missing some stuff in package.keywords. |
Avahi fails to merge because of pygtk is not marked as a needed depency for it. There also isn't use-flag which would add to it --disable-pytgk switch. I think this happens on all platforms but since I am not sure I marked it as AMD64 bug in ebuild. My first bugizilla use ever, so forgive me if something is missing. (I didn't saw need to submit emerge --info) (^^,) "... checking for gdbm.h... yes checking for LIBDAEMON... yes checking for a Python interpreter with version >= 2.4... python checking for python... /usr/local/bin/python checking for python version... 2.4 checking for python platform... linux2 checking for python script directory... ${prefix}/lib/python2.4/site-packages checking for python extension module directory... ${exec_prefix}/lib/python2.4/s ite-packages checking for python module gtk... no configure: error: Could not find Python module gtk !!! Please attach the following file when filing a report to bugs.gentoo.org: !!! /var/tmp/portage/avahi-0.6.10/work/avahi-0.6.10/config.log !!! ERROR: net-dns/avahi-0.6.10 failed. ..." I don't have that config.log file since this can be solved by emerging pygtk.