Tassilo Horn <[email protected]> writes: > David Kastrup <[email protected]> writes: > >> "Tassilo Horn" <[email protected]> writes: >> >>> commit 363b751003c4b425b4b6c511740c95366481207c >>> Author: Luc Van Eycken <[email protected]> >>> Date: Tue Dec 17 09:01:29 2013 +0100 >>> >>> Skip comment or file in parens in output. >>> >>> * tex-buf.el (TeX-parse-error): Skip comment (or file) between >>> parentheses. >>> >>> Signed-off-by: Tassilo Horn <[email protected]> >> >> I'm confused. I was pretty sure that at one point of time I _vastly_ >> simplified TeX-parse-error's recognition of file names to have the >> recognition of opening and closing parens be similarly indiscriminate >> (since detecting spurious ( ) pairings as file names is harmless, but >> skipping some non-filename ( while still recognizing the corresponding ) >> isn't). >> >> Did someone "improve" the file name recognition again? That actually >> makes the problem worse. > > Hm, your last changes to `TeX-parse-error' are from 2006, and thereafter > it's Reiner with one line change in 2007 and Ralf with some changes in > 2009.
Huh. My memory does not match the code. Maybe I confused this with a check that a file actually existed, something which obviously made matching the closing parens worse for the same reasons. The question is how we can improve this. It might make sense to count _all_ opening parens. Or one does not even look at the closing parens at all, but rather at the time of an error throws away bad matches until one gets to something fitting the input. Or one uses the paren count difference as a tie-breaker when more than one element in the list of file names would match the error message. At any rate: making the match condition more selective for opening parens does not help since we can't make the closing paren matches more selective. > Anyway, for the time being, I've reverted the change. > > Luc, could you please explain what's the problem your patch is > actually trying to fix, preferably with a sample file? At least with > the current git version, I don't see a difference with or without your > patch when using the hyperref package. And speaking from the > AUCTeX-date in your patch bug report, your version of > `TeX-parse-error' should be up-to-date. Maybe the idea was a shortcut for discarding matched parens on a single line? In that case, it would not matter whether what we ditch is a file name or not. But it's probably smarter to make a separate match for "paren without good filename". If we see one of those when trying to match actual errors, it's not too late to toss it then rather than not having it considered in the match at all. -- David Kastrup _______________________________________________ auctex-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/auctex-devel
