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

Re: make world - xorg .. CVS source



Mike A. Harris wrote:
Jim Cornette wrote:

After downloading the src rpm and installing the rpm, from the SPEC directory, I then ran the below command on the spec.
rpmbuild -bp --target i686 xorg-x11.spec


Now from the BUILD directory with xorg in, I typed make world w/o editing anything.

The compilation is taking place. What is the output? RPMS or compiled binaries somewhere? Will it be i686 archetecture?


The result will be a compiled X tree, that isn't installed anywhere. The proper way to rebuild for i686, is to take the src.rpm and do:

rpmbuild --rebuild --target i686 <src.rpm>

I noticed that after the very long compilation time that there were no rpms and only the tree.
Thanks for the proper method to create rpms.


Since potitive action has taken place on the 1810 driver, this might be decent to see if the i686 rpms compile and yield any noticable improvements worth the compilation time. There is a patch for the i810, so if not currently included within the build, I might try adding the patch supplied for the build before generating the rpmbuild.

https://freedesktop.org/bugzilla/show_bug.cgi?id=1084

Thanks,

Jim


I must warn however, that rebuilding X with target i686 is totally untested and unsupported, and there are some changes happening right now which might temporarily cause i686 rebuilds to fail. That might occur on Monday's rawhide snapshot. If it does fail (which I hope it doesn't), if you can whip up a patch to fix it and keep i686 ability going, that'd be appreciated. A similar problem might occur for "athlon" rebuilds.




For those interested in why this might occur, it is due to host.def being changed in a fairly major way tonight. From now on, host.def will be a static wrapper file which tests the architecture with C preprocessor defines, and conditionally includes a new file named "host-$arch.def". I've tried to canonicalize all x86 architectures to "i386" so there will be a single host-i386.def file used regardless of wether it is i686, i586, i386, athlon or whatever, but I'm only personally testing to make sure "i386" still works. If others test it and it fails though, patches are graciously accepted to make it work again. ;o)






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