bug#63535: Master branch: Error in forw_comment (syntax.c) handling of escaped LFs

[Alan Mackenzie <acm@muc.de>]

Hello, Eli.

On Tue, May 16, 2023 at 19:29:26 +0300, Eli Zaretskii wrote:
> > Date: Tue, 16 May 2023 16:15:24 +0000
> > Cc: 63535@debbugs.gnu.org
> > From: Alan Mackenzie <acm@muc.de>

> > > >    && !(comment_end_can_be_escaped && char_quoted (from, from_byte))

> > > > ..  Checking char_quoted is wrong.  Instead the function should check the
> > > > current parse state.

> > > Why not both?  IOW, please explain why char_quoted is not TRT here.

> > Because parse-partial-sexp is not scanning the backslash.  The scan
> > starts one character after the backslash, and the syntactic effect of
> > that backslash is not in the OLDSTATE argument to parse-partial-sexp.

> Sorry, I still don't follow: char_quoted doesn't call
> parse-partial-sexp, AFAICT.

parse-partial-sexp calls forw_comment which (wrongly) calls char_quoted.

> So why does it matter what parse-partial-sexp does when we are
> discussing why char_quoted is not TRT?

parse-partial-sexp is the context in which the bug becomes evident.  If,
in the C++ line comment with escaped NL, you start parse-partial-sexp at
the NL, it behaves as though the scan started at the backslash.  This is
the bug.

The cause of the bug is the use of char_quoted at line 42 of
forw_comment.

-- 
Alan Mackenzie (Nuremberg, Germany).