Re: [PATCH 0/6] Fix issues with Python3 scripting

From: Tony Jones
Date: Fri Jan 25 2019 - 13:09:53 EST


On 1/25/19 5:57 AM, Arnaldo Carvalho de Melo wrote:
> Em Fri, Jan 25, 2019 at 01:31:19PM +0100, Arnaldo Carvalho de Melo escreveu:
>> Em Wed, Jan 23, 2019 at 04:52:23PM -0800, Tony Jones escreveu:
>>> Seeteena posted, earlier this week, some patches to add Python3 support
>>> to scripts/python/*.py. Unfortunately there were some issues with these
>>> patches (such as: https://lkml.org/lkml/2019/1/17/351)
>>>
>>> Since I already had a tested set of patches in openSUSE:Factory and
>>> SLE15-SP1 and was about to submit them, Seeteena and I that agreed I
>>> should post my patches not involving scripts/python/*.py and Seeteena
>>> will later resubmit the patches for scripts/python/*.py incorporating
>>> my review feedback under a joint signed-off-by.
>>>
>>> It should be noted that the use of "from __future__ import print_function"
>>> (see: https://lkml.org/lkml/2019/1/16/641) and "except as" (see change to:
>>> tests/attr.py) implies Python2 >= 2.6 as the necessary support has not
>>> been backported to prior versions. I am not sure if it's worth detecting
>>> <2.6 at build time or whether it's sufficiently old as to be a non-issue?
>>>
>>> The shebang changes were driven mostly by our build process as it scans
>>> all files within an rpm and the shebangs would result in a rpm requires
>>> on the python2 binary when BuildRequires was python3-devel. I think they
>>> make sense to apply upstream but understand totally if it's prefered we
>>> keep them local.
>>>
>>> These changes have been tested with PYTHON=python2 (v2.7) and
>>> PYTHON=python3 (v3.6) on latest openSUSE Tumbleweed. I did notice that
>>> test #18 "'import perf' in python" is failing on my system without these
>>> changes. I'll look at it further but didn't want to hold up Seeteena's
>>> resubmit.
>>
>> So it fails on AmazonLinux 1, that has python 2.6, please check if this
>> is something we can workaround, if its difficult, I'll just use
>> NO_PYTHON=1 there to disable it.

Sorry about this. I'll be sure to test more broadly next time.

>>
>> CC /tmp/build/perf/util/parse-branch-options.o
>> util/scripting-engines/trace-event-python.c: In function 'python_start_script':
>> util/scripting-engines/trace-event-python.c:1520:2: error: passing argument 1 of 'PyImport_AppendInittab' discards 'const' qualifier from pointer target type [-Werror]
>> PyImport_AppendInittab("perf_trace_context", initfunc);
>> ^
>> In file included from /usr/include/python2.6/Python.h:130:0,
>> from util/scripting-engines/trace-event-python.c:22:
>> /usr/include/python2.6/import.h:54:17: note: expected 'char *' but argument is of type 'const char *'
>> PyAPI_FUNC(int) PyImport_AppendInittab(char *name, void (*initfunc)(void));
>> ^
>> cc1: all warnings being treated as errors
>> mv: cannot stat '/tmp/build/perf/util/scripting-engines/.trace-event-python.o.tmp': No such file or directory
>> make[5]: *** [/tmp/build/perf/util/scripting-engines/trace-event-python.o] Error 1
>
> I did a quick hack to init an auto variable with that const string and
> then pass it, is passing everything so far:

I see that you already amended in cc4376422552

Thanks!