[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

rpm arch detection vs. configure arch detection



I am having problems with rpmbuild detect architecture differently from 
configure scripts. Its nothing critical really, but makes making generic spec 
files difficult.

On any 686 or better machine, rpmbuild is detecting my hardware as -march=i386 
-mcpu=i686.  This isn't quite right, not a problem in and of it self, but not 
right.  When I am compiling binutils or gcc, these packages make cross 
compiler directories with scripts and hardlinked binaries. rpm is detecting 
the arch as i386, but since configure is detecting as i686 the name of the 
directory is i686-pc-linux-gnu. The issue is that how can we claim ownership 
of these files using macros like %{_target_platform}, when RPM is detecting 
things differently.

Now I know I can just create local .rpmmacros and .rpmrc files for system and 
user level configurations, but this isn't the point.  Shouldn't there be some 
way to have rpm detect by default that the arch is i686 or even athlon ?  
Granted I am speaking stricly within the x86 family of arches and cpus, but 
still....

Further, although a -march=i386 -mcpu=i686 is very reasonable for a 
distribution center creating software packages, it doesn't make very much 
sense for the administrator or developer making packages on a system  by 
system basis.  Again I am not asking the developers to change RPM, just 
asking if there is a way to alter its default behavior.

Oh and as an aside, I have been adjusting the _arch and _target (and the 
otehrs too), macros and I can't seem to get rpm to detect a different arch 
unless I actually define the macro within the specfile.  Any clues?

Thanx for the help,
-- 
Scot Mc Pherson
Linux From Scratch
http://www.linuxfromscratch.org/




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]