GPC compiler crash

Mirsad Goran Todorovac mtodorov at alu.hr
Tue Feb 19 11:37:47 CET 2013


On 19.2.2013. 10:30, Prof A Olowofoyeku (The African Chief) wrote:
> On 18 Feb 2013 at 22:18, Thomas Schneider wrote:
>
>> Garth:
>>
>>> Just to clarify the differences a bit.
>> Thanks for the explanation.  I knew about val/var but wasn't so clear
>> about the heap/stack allocation.  I didn't know that global varibles go
>> on the heap but did know that calling a procedure caused allocation of
>> locals to the stack.
>>
>> So I was making a huge allocation on the stack.  That's fixed now by
>> using a global.  I'm somewhat unclear why the compiler crashed though. I
>> would have thought that if it ran out of stack it would give a warning.
>> But the odd thing is that when the program is being compiled it's not
>> even using the stack yet of course ...
> The compiler crash is a different issue from the program crash. There
> compiler should not crash - in other words, it is a bug. Unfortunately,
> it is a bug that seems unreproduceable elsewhere, and I am not sure
> where that leaves us ...
>

The situation isn't hopeless, as - if having enough time - bug reporter 
(Prof. Schneider) can scale down to the minimal code that still triggers 
compiler bug or crash. If I remember well, this is what Frank and I have 
been doing when bug with SET implementation was triggered only on - then 
rare and now obsoleted - 64-bit Digital Unix on DEC alpha.

They key is to minimize example code that triggers bug, and then it is 
possible to generate test case to prevent bug to reappear. Correct 
implementation of compiler should of course produce correct code on all 
platforms.

Regards,
Mirsad




More information about the Gpc mailing list