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

Re: [Stretch] Status for architecture qualification



On 2016-06-09 02:58, John Paul Adrian Glaubitz wrote:
Hi Alex!


Sorry, i didn't catch this message before the others...

On 06/09/2016 06:44 AM, Alex McWhirter wrote:
If it helps i have access to a decent amount of gear. E6K, V210, V215,
M4000, T1000, T5120, Netra X1, Blade 150, and a V890.

Thanks for the offer. We don't actually lack any hardware, but what we need is new off-the-shelf hardware that would be used to set up the infrastructure for the sparc64 when it becomes an officially supported port in Debian, a
so-called release architecture.

The Debian System Adminstrators (DSA) have the requirement that the hardware they set up for the official buildds and porterboxes is new and not used,
old hardware which might fall apart while being used for critical
infrastructure.

Debian's release architectures come with active security support and for that, the buildds infrastructure needs to be there with high availability to be able to build updated packages with security patches the moment they
become available.

I've been working on the sparc64 port of Gentoo for quite a while. If
there are issues that need addressed for Debian i'm willing to help out
with the effort.

Some unresolved sparc64-related bugs can be found here:

https://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=sparc64;users=debian-sparc@lists.debian.org

You can find more issues if you look at the list of packages with state
"Build Attempted" in the buildd database:

https://buildd.debian.org/status/architecture.php?a=sparc64&suite=sid

Oh, and btw, please upstream any changes you have made in Gentoo to fix
bugs on sparc64. Do you have patches available for the problems with
the testsuites of pulseaudio, util-linux and glibc. Or did you skip
the testsuites for these packages?


Haven't given pulseeaudio a test yet as i haven't really focused on sparc desktops. I do believe i have tests passing on util-linux and glibc, i will have to double check though. These are probably older versions than the unstable debian repository though.

Seems like most of the issues are either bad coding practices or little
mistakes.

True. Another example for that is kdevelop-php:

https://buildd.debian.org/status/package.php?p=kdevelop-php&suite=sid

As you can see, the build fails due to unaligned access:

[  1%] Generating phptokentype.h, phpdebugvisitor.h, phptokentext.h,
phpast.h, phpparser.h, phpparser.cpp, phpvisitor.h, phpvisitor.cpp,
phpdefaultvisitor.h,
phpdefaultvisitor.cpp
cd /<<PKGBUILDDIR>>/obj-sparc64-linux-gnu/parser &&
/usr/bin/kdev-pg-qt --output=php --namespace=Php --debug-visitor
/<<PKGBUILDDIR>>/parser/php.g
Bus error
parser/CMakeFiles/kdev4phpparser.dir/build.make:66: recipe for target
'parser/phptokentype.h' failed
make[3]: *** [parser/phptokentype.h] Error 138

which is usually easy to track down with gdb.

Im sure there's a lot of that amongst the package tree. I haven't worked much on fixing unaligned access type errors, but it's on my todo list.


Thanks for your help!

Adrian


Reply to: