Hi Tassilo,
Le 11/05/2021 à 21:26, Tassilo Horn a écrit :
Hi Jean-Francois,
git bisect indicates first bad commit is
commit 6654955216a42936b87f76dc346aad829b1d52fb
Date: Wed Jun 3 01:44:32 2020 +0900
Use search-based fontification for $...$ (bug#33139)
Reading the commit message, I see there was extensive discussion and
the change seems to have been discussed in depth and well-motivated. I
wonder if you can confirm that you see the same at your locale as me.
Yes, I see the problem. I've pushed a fix which ignores $ in comments
just as it already ignored $ in verbatim contexts.
Could you please test and report back?
Yes, it works also in original context, not only in the short test file.
I realize now that after all I did work for a while with this problem
and I see one file where I have lots of extra $'s added in comments
to create pairs of $'s and avoid the propagation, I removed them
and no ill effect arose, the math highlighting was not applied.
But... wait...
Sadly, I now have another problem. Consider this:
\def\XINT:NE:f:noeval:from:braced:u:p #1#2%
{\detokenize{\romannumeral`$XINT_expr_null\expandafter#1}~expanded{{#2}}}%
There is $ in there, and it triggers math mode highlighting which is a priori
fine,
I can live with that *if I can constrain it to not extend beyond that line*
In the past I would control it this way:
\def\XINT:NE:f:noeval:from:braced:u:p #1#2%
{\detokenize{\romannumeral`$XINT_expr_null\expandafter#1}~expanded{{#2}}}%$
(observe the added $ at end of second line after the % comment character)
With your patch the %$ does not stop the math highlighting !
This is very problematic to me because that problem can't be solved (as I
could solve earlier one, especially now that I understand better what causes it,
via locating extra $'s inside comments)
by adding a commented out $.
It basically means I have not way to stop the highlighting from propagating...