pr21665

[Nick Clifton]

Hi Alan,

> I haven't looked at the bug in detail, but since the testcases are
> 64-bit, is the problem that on a 32-bit target we're not catching a
> size_t overflow?

No - the problem is that the testcase has a pathological .init section:

  % readelf --wide -S POC2
  ...
  [11] .init             PROGBITS        0000000000401ab0 001ab0 800000001a 00  AX  0   0  4
  ...

Note the size - 0x8000000001a - this is too much for xmalloc() to handle,
(at least on my system), and it triggers an error report if run with 
address sanitization enabled.

I do not think that we have to worry about overflow since datasize's type
is bfd_size_type, which is always going to be at least an unsigned long,
right ?

Cheers
  Nick