https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123743

--- Comment #13 from Stefan Schulze Frielinghaus <stefansf at gcc dot gnu.org> 
---
(In reply to [email protected] from comment #11)
> > Am 27.01.2026 um 16:53 schrieb sjames at gcc dot gnu.org 
> > <[email protected]>:
> > 
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123743
> > 
> > --- Comment #10 from Sam James <sjames at gcc dot gnu.org> ---
> > GC then?
> 
> Or address hash?  Does disabling address space randomization help?

Nice catch!  Without ASLR we always get the same object file.

Meanwhile I could reduce it to the following:

$ /usr/bin/s390x-linux-gnu-as --gdwarf-5 -m64 -mzarch -march=z196 -o t1.o
affinity.s
$ /usr/bin/s390x-linux-gnu-as --gdwarf-5 -m64 -mzarch -march=z196 -o t2.o
affinity.s
$ cmp --ignore-initial=16 t1.o t2.o

If you assemble the same input twice you get different results.

Without ASLR we always get

Disassembly of section .rodata.cst8:

0000000000000000 <.LC17>:
   0:   00 00 00 00             .long   0x00000000
   4:   00 0f 1e 78             .long   0x000f1e78

and with ASLR we get different results as e.g.

Disassembly of section .rodata.cst8:

0000000000000000 <.LC17>:
   0:   00 00 00 00             .long   0x00000000
   4:   2f ec                   swr     %f14,%f12
   6:   be 78                   .short  0xbe78

Reply via email to