[Bug gas/16688] aarch64: bignums in literal pools do not work correctly

[will.newton at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16688

--- Comment #1 from Will Newton <will.newton at gmail dot com> ---
ldr q0, =0x5a8279995a8279995a8279995a827999
ldr q1, =0x6ed9eba16ed9eba16ed9eba16ed9eba1
ldr q2, =0x8f1bbcdc8f1bbcdc8f1bbcdc8f1bbcdc
ldr q3, =0xca62c1d6ca62c1d6ca62c1d6ca62c1d6


produces


0000000000000000 <.text>:
   0: 9c000080 ldr q0, 10 <.text+0x10>
   4: 9c0000e1 ldr q1, 20 <.text+0x20>
   8: 9c000142 ldr q2, 30 <.text+0x30>
   c: 9c0001a3 ldr q3, 40 <.text+0x40>
  10: ca62c1d6 .word 0xca62c1d6
  14: ca62c1d6 .word 0xca62c1d6
  18: ca62c1d6 .word 0xca62c1d6
  1c: ca62c1d6 .word 0xca62c1d6
  20: ca62c1d6 .word 0xca62c1d6
  24: ca62c1d6 .word 0xca62c1d6
  28: ca62c1d6 .word 0xca62c1d6
  2c: ca62c1d6 .word 0xca62c1d6
  30: ca62c1d6 .word 0xca62c1d6
  34: ca62c1d6 .word 0xca62c1d6
  38: ca62c1d6 .word 0xca62c1d6
  3c: ca62c1d6 .word 0xca62c1d6
  40: ca62c1d6 .word 0xca62c1d6
  44: ca62c1d6 .word 0xca62c1d6
  48: ca62c1d6 .word 0xca62c1d6
  4c: ca62c1d6 .word 0xca62c1d6

This seems to be caused by the bignum parsing code using a global to store it's
result, but the aarch64 assembler assumes that an expression can be copied and
retain all its state.

-- 
You are receiving this mail because:
You are on the CC list for the bug.