Re: test-strerror_r fails on GNU/Hurd

[Bernhard Voelker]

On 1/3/22 02:52, Paul Eggert wrote:
> On 1/2/22 16:28, Bernhard Voelker wrote:
>> 1073741858: that's 2^30 +34
> 
> Could it be an address, or part of an address, that is incorrectly 
> treated as an integer? A mistake like that is quite plausible given the 
> "does it return char * or int?" API of strerror_r.

no, that number is ERANGE:

  (gdb) p ERANGE
  $4 = 1073741858
  (gdb) p EINVAL
  $12 = 1073741846

and in the 3rd iteration, we get:

  (gdb) p ret
  $8 = 0
  (gdb) p buf
  $9 = "E\000", '^' <repeats 97 times>
  (gdb) p err
  $10 = -3
  (gdb) p i
  $11 = 2

and the following ASSERT aborts the program.

Have a nice day,
Berny