kicad-developers team mailing list archive

[Dick Hollenbeck <dick@xxxxxxxxxxx>]

On 04/04/2014 11:10 AM, Jean-Samuel Reynaud wrote:
> Le 04/04/2014 15:04, Dick Hollenbeck a écrit :
>> I'm sure you are not submitting this for a potential commit, right?
> Clearly the objectif of this patch is NOT to be commit.
>> Because:
>>
>> 1) The comment is below the line of text that it refers to.
> Yes it's a paste an copy from a solution I found on the link:
> http://bugs.python.org/issue19153
>> 2) The comment is bullshit, boost is not part of the discussion.  The comment should read
>> something like, "I don't know why this works, but is solves bug number ???.  It's temporary."
> Yes I forgot to remove the comment when I copy... And of course it's
> temporary to check If the workarround has effect...
> I have try to find a solution by trace the library loading:
> LD_DEBUG=all pcbnew
> But the issue look to be due to a compilation option on some ubuntu
> libs. Anyway I'm NOT sure of that...
>> 3) You are adding python to 7 programs, only one of which is uses python.
> Of course, I repeat this is NOT a patch to commit but a patch to check
> if the issue is the same I found...
>> 4) You are adding python to all operating systems, with no evidence that the problem
>> extends beyond linux.
> Same answer ;)
>> 5) You are adding python to all build, even if python is not used at all, like my builds.
> Same answer...
>> It's basically stupid, without explanation, and no effort was to understand the problem.
>> Its like taking medicine from your Grandmother's medicine cabinet without knowing what it
>> does.  It might have embarrassing consequences.
> As I explain before, it was to identify the issue. If it's fix for
> others users, my idea was to apply this work arround ONLY for the ppa
> building (if I keep the KICAD_SCRIPTING_WXPYTHON option activated).
>> I don't think it should be committed.  I had no idea that wxPython scripting was already
>> so important and widely used.
> On my ppa It's activated. One of the workarround I propose was to
> DISABLE "KICAD_SCRIPTING_WXPYTHON" on ppa. This is the purpose of my
> first email...
> 
>> I think somebody needs to prepare a debugging version of python and single step into the
>> DSO, the call stack at the point of crash passes through wxPyBeginBlockThreads(), and that
>> is the first thing in the libpython.so that is called when called from _pcbnew.kiface.
>> That's what I intend to do, when I get to it.
> Yes it could be great but, as I see some, other software have this issue
> on some Linux distribution.
> I'm NOT sure of that, but as I understand the issue is ouside kicad...

Might be, doesn't matter.  An understanding of the bug is necessary before a wild fix like
this should be applied.  That's just my point of view, as world class software doctor.


Remember that at milestone C) anyone inclined to use python is probably using the python
based project manager, because after all, it will be better than the C++ one given enough
effort.  Under those circumstances, libpython*.so will be loaded at process startup, not
later like now.  Therefore the problem is likely to disappear without effort at that point
in time.


Dick