Re: CVS commit: src/tests/lib/libc/gen

[christos%zoulas.com@localhost (Christos Zoulas)]

On Apr 23,  3:25pm, martin%duskware.de@localhost (Martin Husemann) wrote:
-- Subject: Re: CVS commit: src/tests/lib/libc/gen

| On Mon, Apr 23, 2012 at 09:05:25AM -0400, Christos Zoulas wrote:
| > Keep reading and then it is the following table:
| 
| *sigh*
| 
| And then:
| 
|    For some implementations, the value of si_addr may be inaccurate.
| 
| Can we just bail and use that as an excuse?
| The value on amd64 seems to be completely unrelated to both addresses.
| Anyway, fixing sparc64...

As I mentioned before, the only way to exceed the standard is to
provide implementation specific notes (i.e. document what each
architecture does).  To improve even more, we can provide si_instr
to be the faulting instruction and si_fault to be the faulting
address, and have each architecture fill what is available.

In practice, I doubt that many programs use this feaute.

christos