Re: frysk-core/frysk/proc ChangeLog TestBreakpoint ...

[Andrew Cagney <cagney@redhat.com>]

Going over very very old e-mail.

There was some sort of dead-lock between the code creating the 
event-loop thread, and the code deleting it.  It only occures, and then 
only occasionally, when nothing happens in between.

The way to fix this though, isn't to chase after yet another nasty race 
that is unrelated to what is being unit tested, but rather to look at 
making that test single-threaded - I see that code has suffered from 
other similar nasty races.  Not being multi-threaded means there's one 
less thing that can go wrong.

I also noticed that the test uses file descriptors to exchange messages 
actions rather than signals.  That's a more interesting idea, but does 
mean integrating file-descriptors into the event loop - something I've 
so far avoided.

Andrew

Mark Wielaard wrote:
> Hi Andrew,
>
> On Fri, 2006-10-13 at 09:51 -0400, Andrew Cagney wrote:
>   
>> Mark Wielaard wrote:
>>     
>>> I read that but it doesn't give an explanation for this change.
>>>   
>>>       
>> It stops the hang.
>>     
>
> Yes, it claims that. But why did that hang happen and how does your
> change help? As far as I can see there isn't a change of functionality
> just a duplication of code. So please do explain the change a little
> more. When we understand the cause of the problem we might be able to
> fix it without needing to duplicate code.
>
> Thanks,
>
> Mark
>
>