Unified mouse-highlight code needs testing

Unified mouse-highlight code needs testing

[Eli Zaretskii]

I've committed a few minutes ago to the trunk revision 102261, which
unifies the mouse-highlight code for all sessions, including GUI
sessions and text terminals that support the mouse (GPM and MSDOS).
There's now only one set of note_FOO_highlight functions (in xdisp.c)
and their subroutines that determine which parts of the displayed text
should be highlighted.  The only display-specific part is a single
function which actually draws a given glyph row in mouse-face.

I tested this on MS-Windows (as a representative of GUI sessions) and
on MSDOS.  I don't have access to a system with GPM which actually has
a mouse, so I could only make sure it builds with GPM and runs
correctly without a mouse.  I also made sure the X build compiles and
links.

Please test the new code on a TTY with GPM, on X, and on NS.  On a TTY
with GPM, the immediate bonus from this unification is that now mode
line, header line, and overlays that specify mouse-face should all be
highlighted when the mouse hovers above them.  The easiest way of
testing this is in Info: you have there the mode line, the header
line, and the bread crumbs which all specify mouse-face; these should
all work now.

It is also important to test this in a session that has both X and TTY
frames, especially if the TTY frame uses GPM.

Thanks in advance.

Re: Unified mouse-highlight code needs testing

[Jan Djärv]

Just some points, I'm still compiling.

As termchar.h now is a requrement because of MOUSE_HL_INFO, shouldn't lisp.h 
or or frame.h include it?  nsfont.m didn't and failed to compile (fixed now).

Isn't a name like Mouse_HLInfo explicitly against the GNU coding standards?
Quote: "Stick to lower case; reserve upper case for macros and enum constants, 
and for name-prefixes that follow a uniform convention."

	Jan D.


Eli Zaretskii skrev 2010-11-06 09.56:
> I've committed a few minutes ago to the trunk revision 102261, which
> unifies the mouse-highlight code for all sessions, including GUI
> sessions and text terminals that support the mouse (GPM and MSDOS).
> There's now only one set of note_FOO_highlight functions (in xdisp.c)
> and their subroutines that determine which parts of the displayed text
> should be highlighted.  The only display-specific part is a single
> function which actually draws a given glyph row in mouse-face.
>
> I tested this on MS-Windows (as a representative of GUI sessions) and
> on MSDOS.  I don't have access to a system with GPM which actually has
> a mouse, so I could only make sure it builds with GPM and runs
> correctly without a mouse.  I also made sure the X build compiles and
> links.
>
> Please test the new code on a TTY with GPM, on X, and on NS.  On a TTY
> with GPM, the immediate bonus from this unification is that now mode
> line, header line, and overlays that specify mouse-face should all be
> highlighted when the mouse hovers above them.  The easiest way of
> testing this is in Info: you have there the mode line, the header
> line, and the bread crumbs which all specify mouse-face; these should
> all work now.
>
> It is also important to test this in a session that has both X and TTY
> frames, especially if the TTY frame uses GPM.
>
> Thanks in advance.

Re: Unified mouse-highlight code needs testing

[Eli Zaretskii]

> Date: Sat, 06 Nov 2010 10:38:46 +0100
> From: Jan Djärv <jan.h.d@swipnet.se>
> CC: emacs-devel@gnu.org
> 
> As termchar.h now is a requrement because of MOUSE_HL_INFO, shouldn't lisp.h 
> or or frame.h include it?  nsfont.m didn't and failed to compile (fixed now).

Sorry for missing nsfont.m in my changes.

Did you need termchar.h because MOUSE_HL_INFO references
output_data.tty->display_info, or is there another issue?

As for your question, I don't think we require that a header be
included by another header if they always need to go together.  We
already have other cases like this one: for example, termhooks.h uses
FRAME_X_P and FRAME_W32_P, but doesn't include frame.h where these are
defined.

That said, I have no objections to include termchar in frame.h, of
course, provided that it doesn't have unintended consequences.

> Isn't a name like Mouse_HLInfo explicitly against the GNU coding standards?

I had Display_Info as the model, but I can rename the typedef, of
course, if Stefan and Chong don't like this name.

Re: Unified mouse-highlight code needs testing

[Jan D.]

  Eli Zaretskii skrev 2010-11-06 11:41:
>> Date: Sat, 06 Nov 2010 10:38:46 +0100
>> From: Jan Djärv<jan.h.d@swipnet.se>
>> CC: emacs-devel@gnu.org
>>
>> As termchar.h now is a requrement because of MOUSE_HL_INFO, shouldn't lisp.h
>> or or frame.h include it?  nsfont.m didn't and failed to compile (fixed now).
> Sorry for missing nsfont.m in my changes.
>
> Did you need termchar.h because MOUSE_HL_INFO references
> output_data.tty->display_info, or is there another issue?

No, thats it.

> As for your question, I don't think we require that a header be
> included by another header if they always need to go together.  We
> already have other cases like this one: for example, termhooks.h uses
> FRAME_X_P and FRAME_W32_P, but doesn't include frame.h where these are
> defined.
>

Ok.

     Jan D.

Re: Unified mouse-highlight code needs testing

[MON KEY]

> I've committed a few minutes ago to the trunk revision 102261, which
> unifies the mouse-highlight code for all sessions, including GUI
> sessions and text terminals that support the mouse (GPM and MSDOS).

I'm consistently crashing Emacs built from trunk @ rev 102388 whenever
I mouse over regions with mouse highlighting face.

I've no idea if this is related but it seems so.
Unfortunately I'm unable to supply a proper backtrace right now.

--
/s_P\

Re: Unified mouse-highlight code needs testing

[Eli Zaretskii]

> Date: Mon, 15 Nov 2010 01:36:39 -0500
> From: MON KEY <monkey@sandpframing.com>
> 
> > I've committed a few minutes ago to the trunk revision 102261, which
> > unifies the mouse-highlight code for all sessions, including GUI
> > sessions and text terminals that support the mouse (GPM and MSDOS).
> 
> I'm consistently crashing Emacs built from trunk @ rev 102388 whenever
> I mouse over regions with mouse highlighting face.

Ouch!

> I've no idea if this is related but it seems so.
> Unfortunately I'm unable to supply a proper backtrace right now.

Can you at least show a recipe for reproducing this starting with
"emacs -Q"?

Re: Unified mouse-highlight code needs testing

[Eli Zaretskii]

> Date: Mon, 15 Nov 2010 18:14:35 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> > Date: Mon, 15 Nov 2010 01:36:39 -0500
> > From: MON KEY <monkey@sandpframing.com>
> > 
> > > I've committed a few minutes ago to the trunk revision 102261, which
> > > unifies the mouse-highlight code for all sessions, including GUI
> > > sessions and text terminals that support the mouse (GPM and MSDOS).
> > 
> > I'm consistently crashing Emacs built from trunk @ rev 102388 whenever
> > I mouse over regions with mouse highlighting face.
> 
> Ouch!
> 
> > I've no idea if this is related but it seems so.
> > Unfortunately I'm unable to supply a proper backtrace right now.
> 
> Can you at least show a recipe for reproducing this starting with
> "emacs -Q"?

Also, is that a GUI session that crashes, or a TTY with GPM?