Re: Lots of skips on python3[78].

From: Stefan Esser <se_at_freebsd.org>
Date: Fri, 31 Dec 2021 21:47:33 UTC 
Am 31.12.21 um 20:20 schrieb Larry Rosenman:
> On 12/31/2021 4:07 am, Stefan Esser wrote:
>> Am 30.12.21 um 19:47 schrieb Larry Rosenman:
>>> https://home.lerctr.org:8888/build.html?mastername=live-host-ports&build=2021-12-30_11h08m50s
>>>
>>>
>>> Can someone look?
>>
>> This is fall-out from my commit aiming to reduce incompatibilities
>> of macros defined in sys/cpuset.h with those expected by programs
>> written with GLIBC in mind.
>>
>> Python needs additional macros defined in GLIBC, and I'm going to
>> add them after a buildworld/buildkernel has succeeded.
>>
>> I have already verified that Python builds succeed with this planned
>> commit.
>>
>> Sorry for the breakage - I had tested the changes with a large number
>> of locally built ports, but missed the build issue in the Python port.
>>
>> Regards, STefan
> 
> Stefan,
>    Can you look at
> https://home.lerctr.org:8888/data/live-host-ports/2021-12-31_13h11m13s/logs/errors/turbostat-4.17_2.log
> 
> It's got CPU_* macro issues, and is unmaintained.

Hi Larry,

I have updated the port to build on -CURRENT (commit 6cc600ed7040).

It did define the macros that have been added to sys/cpuset.h, and
the duplicate definitions caused the compilation to fail.

But the port has functional issues. The variable "cpu_subset" in
topology_probe() is only initialized to a non-NULL pointer if the
program is invoked with a -c option, for example.

The port is based on 4 year old upstream sources, which have last
been updated in May of this year.

I'm not going to fix the functional issues of this port now, but
might look into it at a later time.

Regards, STefan