Hppa distribution

From openSUSE

The SUSE hppa distribution is just work in progress. There is nothing you would like to install unless you are a real expert and used to doing everything manually. There is no yast, no hwinfo; installation needs to be done without the help of tools. In other words, you don't want to have it.

There are six machines running and building this distribution currently:

• one C3750
• one C3700
• one C3650
• two C3600
• one C3000

Major work has to be done on several main packages:

• glibc builds again. However, the testcases that fail a different than those upstream, even in the same build environment. Looks like the SUSE glibc fixed some bugs, but made some other problems for hppa. However, this is good progress.
• kernel does exist as a package. Currently blogd blocks the machine from coming up. I workaround this problem by disabling blogd in the initrd. The parisc linux kernel patches are very close to mainline. That makes things a lot easier for me (many thanks to the people that do this, especially Kyle McMartin for maintaining the special git tree for parisc).
• Having no -fstack-protector makes some packages fail that also want to build with -Werr.
• yast never could be compiled, because hwinfo does not compile. Last time I did that manually, calling hwinfo crashed the machine. Another problem with building yast is: cc1plus: out of memory allocating 4072 bytes after a total of 640262144
• a minor problem where I simply have to update the texlive package is, that in our build system, ulimit -s unlimited fails for hppa. This should be changed to 81920, but unfortunately, the noarch package does not hit the rule that I already checked in.
• one major problem right now is, that linking with -pie currently creates only binaries that segault when running them. I still hope that this is resolved somehow when I manage to build glibc again. UPDATE: this seems to be fixed with the latest glibc (it is only submitted, not yet built in autobuild), what means I still have to check.
• I provided a mpfr package manually, because that is essential to build. However, running the mpfr tests shows that this is totally broken. 130 of 133 tests simply fail. Maybe the situation will improve with the new glibc.