bug#23461: perl-mode: Displaying HERE-docs as strings instead of comments [PATCH]

[Harald Jörg]

This is a detour from my work on CPerl mode bugs ... while trying to
steal some syntax concepts from perl-mode.pl, I stumbled over this old
report.

I guess I can explain what's going on.  Short story: Perl mode marks
HERE-docs syntactically as c-style comments, hence font-lock-mode
selects the comments face.

Investigating how to fix this leads to the longer story.  There are two
possible approaches:

  1) use a string-style syntax (generic string) instead of c-style
     comments to flag HERE-documents.  That way, font-lock picks up the
     correct face automagically.

  2) Keep HERE_docs as c-style comments, but change the face mapping by
     injecting a function into font-lock-defaults which applies the
     string face to c-style comments.

Both approaches work, but both are a bit whacky.  For 1), changing the
syntax code is easy but opens a can of worms: Indentation after the
HERE-doc doesn't work any more.  The reason is that Perl mode needs to
go "back" to find out whether a statement is a continuation line.
"Back" includes skipping back over comments, but that HERE-doc is no
longer a comment, so it blocks the way to find whether the line before
the HERE-doc ends a statement.  To fix that, all calls to
perl-backward-to-noncomment must be checked whether they need to skip
backward over HERE-docs, too.  I added a function
perl-backward-to-noncomment-nonhere, and eventually it turned out that
the simple perl-backward-to-noncomment seems to be superfluous.

For 2), it feels wrong to have strings marked as comments, and it is a
bit of a hack to insert a function into font-lock-keywords which doesn't
even search for keywords.  CPerl mode uses a similar trick, but CPerl
mode is renowned for being whacky.  Also, __DATA__ sections in Perl mode
are marked generic strings, so there ought to be some disambiguation.

The patch uses the first approach, and also adds tests which are
independent of the chosen solution.


As a bycatch, it also fixes the case where the line starting a HERE-doc
ends in a comment, which was messed up by perl-mode.  I could not find a
bug report for that but test cases are included.

Perhaps Stefan has an opinion on this, and chances are good that he can
point to a better solution...
-- 
Happy winter solstice,
haj

0001-perl-mode-Display-here-docs-as-strings-instead-of-co.patch
Description: perl-mode: Treat HERE-docs as strings