Re: MacOS X problem with the gui

[Alexander Hansen]

On 8/24/12 7:36 PM, Daniel J Sebald wrote:
> On 08/24/2012 08:55 PM, Ben Abbott wrote:
>>
>> On Aug 24, 2012, at 11:46 AM, Ben Abbott wrote:
>>
> 
>>> Break on
>>> __THE_PROCESS_HAS_FORKED_AND_YOU_CANNOT_USE_THIS_COREFOUNDATION_FUNCTIONALITY___YOU_MUST_EXEC__()
>>> to debug.
>>>
>>> A description of what can cause these errors is at the link below.
>>>
>>> http://objectivistc.tumblr.com/post/16187948939/you-must-exec-a-core-foundation-fork-safety-tale
>>>
>>>
>>> Ben
>>>
>>> <octave_2012-08-24-114342_Bens-MacBook-Pro.crash>
>>>
>>
>> Same problem had been reported for gnuplot.  Anyone know how to look
>> track down the changeset?
>>
>> http://sourceforge.net/tracker/index.php?func=detail&aid=3469233&group_id=2055&atid=102055
>>
> 
> The bug report says it was closed because of being out of date.  That's
> not real helpful, meaning there likely wasn't anything done to fix it
> and it went away on its own.  OR, the code is set up so that if the fork
> fails it attempts something without a fork.
> 
> There aren't any comments associated with changesets.  gnuplot maintains
> a ChangeLog of exactly what has been modified, and they are usually
> descriptive enough.  I looked and saw only one comment about fork() on
> 2012-07-05.  One would then use the date to browse the diff in
> CVS/Sourceforge.  But don't do that in this case because the comment
> didn't relate to what you've found.  I will ask Mojca why that bug
> report was resolved.
> 
> Are the comments you are seeing:
> 
> __THE_PROCESS_HAS_FORKED_AND_YOU_CANNOT_USE_THIS_COREFOUNDATION_FUNCTIONALITY___YOU_MUST_EXEC__()
> 
> 
> actual errors?  Or a limitation of your debugger?  It looks to me like
> the debugger is telling you it can't follow forks, which understandably
> might be kind of tricky because the debugger has this whole history of
> things it has done and variables it has tracked and that would mean
> replicating all that stuff.  But, exec() to the debugger might look like
> just starting an additional debug session (i.e., from scratch).
> 
> Perhaps the resolution is to not turn on debug code.
> 
> But, this issue with fork() and Qt is sort of what I'm wondering about
> Windows.  That would help gnuplot developers as this fork() in Qt might
> not be essential; it was patterned after the X11 terminal without
> thought of what fork() is for in this case.
> 
> Dan

It's an actual error, no debugger required.

You'll get this on the console if you fork() and your executable links
to CoreFoundation.framework, or even if your executable links to
something else that itself links to CoreFoundation, etc.  It became a
fatal error for OS 10.5 and later.

-- 
Alexander Hansen, Ph.D.
Fink User Liaison
My package updates: http://finkakh.wordpress.com/