Bugzilla Bug 50616

In the nessus GUI application or xmms only tabs are shown in the windows, and
no buttons or labels are showing up.

Reproducible: Always
Steps to Reproduce:
1. Emerge xmms
2. Start xmms
3. Press CTRL-P to bring up the preferences-dialog

Actual Results:  
Preference dialog came up, but it missed all the buttons.

Expected Results:  
Expected the buttons and labels to show up as normal.

Removing -msse and -msse2 from CFLAGS in /etc/make.conf and re-emerging
gtk+-1.2.10-r11 fixes the problem.

cflags are the users responsibility.

CFLAGS which break the build should be masked by the ebuild. AFAIK, the
openoffice ebuild does this. I would not consider -msse and -msse2 as
experimental CFLAGS. To get normal Pentium-M optimizations one need to use
'-march=pentium3 -msse2' as CFLAG to get Centrino features (until
-march=pentium-m is added to gcc) isn't part of gcc yet. At least the GTK
1.2.10-r11 ebuild should state with big letters that sse/sse2 optimizations
will break functionality of the package.

it's too much a gcc/glibc/whatever thing. There's always 1 or 2 users who have
a certain problem with certain cflags in certain circumstances, should we cap
the cflags for all users for that ? No, we don't add caps until there's been a
lot of reproducable proof on different systems.

Considering this ebuild & has been around an eternity and you are the first to
report such problems, I don't consider it evidence enough this is more than a
local glitch.

Well. Unless GCC fix the SSE2 optimizations we will probably be bitten by this
bug once people start using -march=pentium-m in their CFLAGS.

According to http://gcc.gnu.org/ml/gcc-patches/2004-02/msg01328.html the only
difference from pentium3 and pentium-m is SSE2 (at least to GCC).

At least consider putting it into the documentation/faq/howto somewhere.
Centrino laptops are getting more and more common... BTW: This was on a Dell
Precision M60.

my make.conf has a warning about using pentium-4 and sse2 . I guess that should
be extended to the pentium-m then.

I wonder why we don't disable those flags by default for those architectures if
we know they don't work.

gcc team, please check out comment #4 & #5

how about this:
'dont use -msse2 with pentium-m unless using gcc-3.4 or later' ?

you never mentioned what version of gcc you're using so i'm going to have to assume 3.3.{2,3} in which case i say we do nothing to 'fix' this bug except tell the users of pentium-m the above warning

I'm using the gcc-3.3.2-r5, so you're quite correct.

A warning like that would be most beneficial.

BTW: When is gcc 3.4 slated for stable release in gentoo? 

3.3.3-r4 is probably the next stable version of gcc to happen ... we're still
working out different things with gcc-3.4 and until that settles, that wont be
in stable

nick: can you update make.conf to say something like

Do not use -msse2 with Pentium-M CPU's unless using gcc-3.4 or higher.

Noted for future make.conf's. Not a portage bug, reassigning.

should be set ... not much more we can do