Re: Large size array problem

[Cserzo Miklos]

On Thu, 19 Jun 2003, Nick Clifton wrote:

Hi Nick,

> Please could you try the attached patch and let me know if it works
> for you.  At the moment it only corrects the sparc target, but I can
> extend it into a generic fix if it appears to be the correct approach.

I made the tests and it is negative. The 'gas' compiles and installs fine.
The 'gcc' and 'gpc' compiles and installs too with the 64 bit patched
'gas'. The code that was OK with the old version is fine with the new one
too and runs OK. The code that made the old version fail compiles now and
produce a valid 64 bit executable according to 'file'. The worrying signs:
I try to allocate a vector of size 1 billion long-word (8 Gig). It
compiles, start to run but 'top' sees only less than 4 Gig memory
allocated for the process, than core-dump generated after a while due to
segmentation fault. So it looks like the requested memory is not
actually allocated and the process goes tits up trying to access it.

Cheers,

miklos