Re: [PATCH] libctf: check for problems with error returns

[Nick Alcock <nick.alcock@oracle.com>]

On 16 Oct 2023, Torbjorn SVENSSON told this:
> On 2023-10-15 21:18, Nick Alcock wrote:
>>>> +  if (ctf_member_info (fp, stype, "bar", &mi) < 0)
>>>> +    fprintf (stderr, "cannot get member info: %s\n", ctf_errmsg (ctf_errno (fp)));
>>>> +
>>>> +  /* Iteration should never produce an offset bigger than the offset just returned,
>>>> +     and should quickly terminate.  */
>>>> +
>>>> +  while ((ret = ctf_member_next (fp, stype, &i, NULL, NULL, 0)) >= 0) {
>>>> +    if (ret > mi.ctm_offset)
>>>> +      fprintf (stderr, "ssize_t return: unexpected offset: %zi\n", ret);
>> (here.)
>
> Ah, okay. In any case, I think it would be clearer if you get all the lines in one go that fails than just the first one (in case of
> multiple failures...). - But, that's only my 2 cents.

We do that for failures that actually relate to what's being tested, but
these lines are only setup, and each depends on the one before: if any
fail, all the ones after it are certain to.

I think I might change it to not do the actual tests if we can't do the
setup, though!

I'll give v8 a quick sanity check: more in about an hour.

-- 
NULL && (void)