Eric Blake wrote:
> On 01/27/2011 08:16 AM, Jim Meyering wrote:
>>>>> @grep -nEA2 '[^rp]error *\(' $$($(VC_LIST_EXCEPT)) \
>>>
>>> Is there a way to tighten it, by passing the grep output through a sed
>>> script that combines multiple lines (easily determined by the difference
>>> between file: and file- for matching vs. context lines and the --
>>> separator between hunks), then limit the detection of the capital letter
>>> to one occurring after error( but before ';'?
>>
>> That sounds like it'd be an improvement. With that, we could
>> search a couple more lines of following context, too.
>> While I haven't been willing to spend any more time on it,
>> you're welcome to do so.
>>
>> One tiny fly in the ointment: ';' inside a string or in a comment
>> between concatenated strings would cause trouble.
>
> Yes, it means possible false negatives, but that's probably not a
> practical concern. Really, who writes code this awful, with a ; in a
> comment prior to the error message?
>
> error (0, errno, /* ; */ _("Oops"));
It doesn't have to be awful:
error (0, errno, /* a few words; more words. */
_("Oops"));
but, as I said, it's a tiny fly.
> Whereas a ';' inside a string is not too likely to cause issues; this
> (mis-)usage would still be caught:
>
> error (0, errno, "%s", _("Oops; you did something wrong"));
>
> because we are looking for any instance of "[A-Z] after the open ( but
> before the first ;, and a ; within the string is after the point we care
> about.
Good point.
