[mythtv] CVS broken for -march=athlonxp

Jesper Sörensen jesper at datapartner.se
Mon Nov 15 23:14:10 UTC 2004

Joseph A. Caputo wrote:

>On Monday 15 November 2004 16:25, Jesper Sörensen wrote:
>>Compiling with -march=pentiumpro instead of -march=athlonxp seems to
>>fix it but athlonxp always worked before so it would be nice if it
>>could be fixed again. :)
>>g++ (GCC) 3.3.4 (Debian 1:3.3.4-13)
>>/usr/src/linux is 2.6.10-rc2
>>With "CONFIG += release":
>>make[3]: Entering directory
>>g++ -c -pipe -w -O3 -march=athlonxp -fomit-frame-pointer -fPIC -DPIC
>>-fomit-frame-pointer -D_REENTRANT  -D_GNU_SOURCE
>>-I/usr/share/qt3/mkspecs/default -I. -I/usr/include
>>-I/usr/local/include -I/usr/src/linux/include -I../..
>>-I/usr/include/qt3 -o AAFilter.o AAFilter.cpp
>>cc1plus: error: bad value (athlonxp) for -march= switch
>>cc1plus: error: bad value (athlonxp) for -mcpu= switch
>>make[3]: *** [AAFilter.o] Error 1
>(1) this has nothing to do with Myth.  The error is coming from the 
>compiler itself, and it's telling you that you've passed an invalid 
>command line option (athlonxp) to the "-march" or "-mcpu" switch.
>(2) If you looked at the GCC(1) man page, you'd see that the reason it's 
>complaining is that the option is "athlon-xp", not "athlonxp"

You're absolutely right! I don't know where I got the athlonxp from, but 
that's what I've been using when compiling Myth the last year or so. 
Funny thing it wasn't until the soundtouch fix it started complaining 
(compiled just fine two days ago). I guess it's always been silently 
ignoring that swith or something.

>(3) based on previous posts to the lists, "-march=athlon-xp" gives worse 
>performance than "-march=pentiumpro", even on an Athlon XP.

Not what I would have expected but I'll take your word for it. Thanks 
alot for you help! Everything is working again now. =)

More information about the mythtv-dev mailing list