On Wed, Dec 30, 2009 at 06:07:13PM +0100, Thomas Sailer wrote:
> Yeah and I missed one too :)

Hi,

thank you again.

> There's a call to snprintf in Grt.Rtis_Utils.Get_Value. It currently
> works by sheer luck, because Type_Rti.Kind ends up in %eax, then gcc
> subtracts 0x16 from %eax, and then it calls snprintf. This means at the
> call to snprintf, %al contains 5, which is in the allowed range and at
> least 1 to copy the used %xmm0 register to the vararg structure.
> 
> But should the register allocation order or the exact jump table
> implementation ever change in gcc, we risk segfaults.

Sure, this works by luck.

> Would be nice if you could check this in as well.

Will do.

> Thanks and have a nice new year's eve!

Thank you for tracking this and happy new year too.

Tristan.

_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss

Reply via email to