[ft-devel] FW: Stack Overflow ?

[Chuck Burfoot]

Have been reading DrMemory. However, this isnt' unusual for MS release builds 
in particular.

As I mentioned and you're obviously aware, why DON'T you used calloc internally?
This is running rock solid with that 1 change in ft_alloc()

I consider this fixed until/unless something else goes boom (shouldn't be set 
to 0 by default)
But this runs solid in debug and now release with calloc in ft_alloc()


-----Original Message-----
From: address@hidden [mailto:address@hidden On Behalf Of Werner LEMBERG
Sent: Saturday, March 22, 2014 10:12 AM
To: address@hidden
Subject: Re: [ft-devel] Stack Overflow ?


> So I analyzed how the memory was allocated. I realized that EVERYTHING 
> was alloced from ft_alloc(memory, size) so I substituted
> ft_scalloc(size,1) for ft_smalloc ...  Presto bingo - everything seems 
> to be JUST FINE

This essentially only hides a bug – if there is a bug.

> It's the Microsoft (and sometimes Linux or Mac) uninitialized memory 
> error.

FreeType has been stress-tested a lot with very picky compilers that find 
uninitialized memory issues at runtime.  No issues are known currently.  I 
*really* doubt that this is an issue here; it must be something else.

> THIS is one of the dirtiest issues in release builds with Visual 
> Studio - indeterminate state of allocated memory AND can be especially 
> insidious when it's a variable on the stack

What about using Dr. Memory to get more information?  I've never tried it, but 
it seems to be an equivalent to `valgrind' on GNU/Linux...

   http://drmemory.org


     Werner