Re^3: Different behaviour of 'stat' function after assignment to $! (errno).

[Anonymous Monk]

Nothing "funny" has been done to $! before the second Dump, but it hasn't changed since before the stat.

$! , errno, is supposed to be positive, so using -$! is something funny :)

[ikegami]

Negating positive numbers isn't doing something funny.

[jpl]

$! , errno, is supposed to be positive, so using -$! is something funny :)

All the "funny" stuff happens after the second Dump, which happens immediately after the stat. Unless the optimizer is punishing us for something that is going to happen later, I cannot explain why the failing stat did not appear to modify $!.

[Anonymous Monk]

It does modify $!, but its only reflected when you Peek if you put $! in int/boolean/string context first, because Peek won't force the conversion. The assignment modified the IV slot, but stat modified the NV/PV slot.

$ perl -MDevel::Peek -E " stat 1; Dump($!); $!=666; stat 1; $! and Dum
+p($!); "
SV = PVMG(0x9b3bfc) at 0x9ba5ec
  REFCNT = 1
  FLAGS = (GMG,SMG)
  IV = 0
  NV = 0
  PV = 0
  MAGIC = 0x9d904c
    MG_VIRTUAL = &PL_vtbl_sv
    MG_TYPE = PERL_MAGIC_sv(\0)
    MG_OBJ = 0x9ba61c
    MG_LEN = 1
    MG_PTR = 0x9da014 "!"
SV = PVMG(0x9b3bfc) at 0x9ba5ec
  REFCNT = 2
  FLAGS = (GMG,SMG,pNOK,pPOK)
  IV = 666
  NV = 2
  PV = 0x3ffee4 "No such file or directory"\0
  CUR = 25
  LEN = 28
  MAGIC = 0x9d904c
    MG_VIRTUAL = &PL_vtbl_sv
    MG_TYPE = PERL_MAGIC_sv(\0)
    MG_OBJ = 0x9ba61c
    MG_LEN = 1
    MG_PTR = 0x9da014 "!"