Stack overrun in JIT

[classilla]

Regression range is FPR6 (ok in FPR5). Uplifted from https://tenfourfox.tenderapp.com/discussions/problems/8587-tenfour-fox-r6-quits and https://tenfourfox.tenderapp.com/discussions/problems/8592-deviant-art

Test sites include https://deviantart.com/users/login and https://www.boxed.com/login/

[classilla]

The trace invariably includes this terminal stack address:

Program received signal EXC_BAD_ACCESS, Could not access memory.
Reason: KERN_INVALID_ADDRESS at address: 0xbf7ffff0

But that address is unpossible unless somehow our stack address and size weren't being incorporated.

[classilla]

... and they aren't. In FPR5, the stack size appears correct:

Load command 3
      cmd LC_SEGMENT
  cmdsize 56
  segname __UNIXSTACK
   vmaddr 0xb0000000
   vmsize 0x40000000
  fileoff 20480
 filesize 0
  maxprot rwx
 initprot rw-
   nsects 0
    flags (none)

In FPR6, this load command doesn't even exist. The only changeset in this range that affects that is 74793a8.

[classilla]

@rmottola You're busted: backing that commit out fixes the problem and the stack size is correctly placed. I'm trying to see if I can determine why it went wrong, or if it has to backed out fully.

[rmottola]

if CONFIG['OS_ARCH'] == 'Darwin' and CONFIG['CPU_TYPE'] == 'ppc':

seems really pretty safe, is there some syntax issue and/or OS_ARCH and CPU_TYPE are wrong? They are not used in the same file, I copied them from elsewhere.

It is important, since after your suggestion I did the same in jit, fixing JIT crashes on PPC

[classilla]

CPU_TYPE should be CPU_ARCH.

Can you set up a PPC build environment to test future changes? This issue has the potential to break a lot of sites and I'm mulling over whether I need to issue an emergency update.

[classilla]

This is fixed and tests properly, so closing.