Hello,
I no longer have easy access to a PowerPC machine, and won’t be able to reproduce issues for it. I’ll step down as the PEP-11powerpc64le-unknown-linux-gnu (GCC) contact.
Unless someone else steps up, that will drop powerpc64le to Tier 3, meaning that failures don’t block a release. It will leave @vstinner as the contact for both gcc and clang builds on the architecture.
To be clear, I’ll be happy to help with issues on this and other underrepresented architectures – IMO, that helps CPython’s overall portability. But I won’t be able to fix issues myself.
If you’d like to test Python on it, go right ahead, but there’s probably not much use in providing official support.
powerpc64le is a different platform. AFAIK, nowadays it’s mainly used in high-end supercomputer clusters. It’s not easy to get access to such a system.
If someone is interested to help maintaining this platform, I can provide them guidance on how to get a development machine running ppc64le. For example, I can provide a SSH access to ppc64le machine operated by Red Hat. Please contact me in private if you are interested.
In my experience, there are only minor differences between x86-64 and ppc64le regarding Python development, so it’s not a big burden to support this platform. For example, one difference is that ppc64le use memory pages of 64 kB instead of 4 kB and some code makes the assumptions that a page is always 4 kB. Or recently, test_gdb failed more often on this platform: gdb is closer to machine code and so more likely to deserve more love to get it working on more platforms.
The main issue is to get access to a ppc64le machine. As I wrote, I can help you for that
Thanks!
I definitely expect to use your help; you’re a great Contact person for the architecture!
When I signed up I had direct chat to ppc experts. Now, I’d have to go through you, and if you’d be on vacation I couldn’t meet the 24-hour target SLA.
FWIW, I also expect to spend more time on buildbot failures in general, so I might actually pay more attention to ppc than before. Just not as a priority
test_gdb failed more often on this platform
Yeah, I don’t really consider gdb failures to be breaking the platform.