Re: [AUCTeX-devel] New error parsing

[Mosè Giordano]

Hi Tassilo,

I noticed that the regexp in `TeX-parse-error' doesn't match vertical
bad boxes messages, that usually are like

  Underfull \vbox (badness 1048) has occurred while \output is active [7]

where [7] is the page where the bad box occurred, if I got it right.
Ok, this message is pretty useless as it is because it doesn't provide
the offending file and line, but nevertheless I think we should catch
it.

In addition, in a document of mine I have some bad horizontal boxes
with messages like

  Overfull \hbox (0.93071pt too wide) detected at line 29

but the regexp expects it to end with "at lines 12--34".  How about
replacing the regexp for horizontal bad boxes with

--8<---------------cut here---------------start------------->8---
"^\\(\\(?:Overfull\\|Underfull\\|Tight\\|Loose\\)\
 \\\\.*?lines? [0-9]+\\(?:--[0-9]+\\)\\)"
--8<---------------cut here---------------end--------------->8---

?  I'm not sure, though, the phrase "detected at line(s)" always
appears in such warnings.  As an alternative:

--8<---------------cut here---------------start------------->8---
"^\\(\\(?:Overfull\\|Underfull\\|Tight\\|Loose\\)\
 \\\\.*?[0-9]+\\(?:--[0-9]+\\)?$\\)"
--8<---------------cut here---------------end--------------->8---

with the trailing "$".  What do you think?

Bye,
Mosè


2016-02-11 18:43 GMT+01:00 Tassilo Horn <address@hidden>:
> Mosè Giordano <address@hidden> writes:
>
>> First bug spotted!
>
> Oh, that's been quicker than I've thought. :-)
>
>> Sometimes after the file name there are the pages of the output file:
>>
>>     (./path/to/file [8] [9] [10]
>>
>> The complex regexp of the old parsing function was able to ignore the
>> page numbers, instead the new implementation captures the numbers as
>> well.  In order to keep the regexp as simple as possible,
>> TeX-parse-error now removes from the string leading and trailing
>> whitespace and these numbers.  Feel free to suggest better solutions
>> for this problem.
>
> No, I think that's fine.
>
>> However, now there is also an ERT test for error parsing, every bug
>> fixed in this field should (read: must) come with a proper test, in
>> order to make sure we don't break anything.
>
> Oh, yeah, that's a good idea!
>
> Bye,
> Tassilo
>
>
> _______________________________________________
> auctex-devel mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/auctex-devel