bug#10960: linum-mode does not work properly when changing font size

bug#10960: linum-mode does not work properly when changing font size

[Le Wang]

[-- Attachment #1: Type: text/plain, Size: 343 bytes --]

Here is the stackoverflow question:

http://stackoverflow.com/questions/9304192/emacs-linum-mode-and-size-of-font-unreadable-line-numbers

Here is another question with a hacky solution

http://unix.stackexchange.com/questions/29786/font-size-issues-with-emacs-in-linum-mode/30087#30087

I can repro this with recent builds of Emacs24

-- 
Le

[-- Attachment #2: Type: text/html, Size: 626 bytes --]

bug#10960: linum-mode does not work properly when changing font size

[Glenn Morris]

Le Wang wrote:

> Here is the stackoverflow question:
>
> http://stackoverflow.com/questions/9304192/emacs-linum-mode-and-size-of-font-unreadable-line-numbers

Here is the existing Emacs report

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=1255

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Date: Tue, 06 Mar 2012 12:07:37 -0500
> Cc: 10960@debbugs.gnu.org
> 
> Le Wang wrote:
> 
> > Here is the stackoverflow question:
> >
> > http://stackoverflow.com/questions/9304192/emacs-linum-mode-and-size-of-font-unreadable-line-numbers
> 
> Here is the existing Emacs report
> 
> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=1255

I don't understand Stefan's response in that bug: what effect exactly
are text-scale-* functions supposed to produce?  Should the text in
the margins be enlarged to the same size as the text in the "text
area" of the window, or should the text in the margins retain its
original size?

And whichever of these two is TRT, why is this considered to be a
problem in linum-mode, and not in the infrastructure (a.k.a. the
display engine)?

bug#10960: linum-mode does not work properly when changing font size

[Juanma Barranquero]

On Tue, Mar 6, 2012 at 19:11, Eli Zaretskii <eliz@gnu.org> wrote:

> I don't understand Stefan's response in that bug

I didn't either at the time, nor do I now...

    Juanma

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

Ping!  Stefan, could you please explain your response?

> Date: Tue, 06 Mar 2012 20:11:43 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 10960@debbugs.gnu.org, l26wang@gmail.com
> 
> > From: Glenn Morris <rgm@gnu.org>
> > Date: Tue, 06 Mar 2012 12:07:37 -0500
> > Cc: 10960@debbugs.gnu.org
> > 
> > Le Wang wrote:
> > 
> > > Here is the stackoverflow question:
> > >
> > > http://stackoverflow.com/questions/9304192/emacs-linum-mode-and-size-of-font-unreadable-line-numbers
> > 
> > Here is the existing Emacs report
> > 
> > http://debbugs.gnu.org/cgi/bugreport.cgi?bug=1255
> 
> I don't understand Stefan's response in that bug: what effect exactly
> are text-scale-* functions supposed to produce?  Should the text in
> the margins be enlarged to the same size as the text in the "text
> area" of the window, or should the text in the margins retain its
> original size?
> 
> And whichever of these two is TRT, why is this considered to be a
> problem in linum-mode, and not in the infrastructure (a.k.a. the
> display engine)?
> 
> 
> 
> 

bug#10960: linum-mode does not work properly when changing font size

[Stefan Monnier]

> Ping!  Stefan, could you please explain your response?

I'm not sure what to explain about it.

>> I don't understand Stefan's response in that bug: what effect exactly
>> are text-scale-* functions supposed to produce?  Should the text in

Increase the size of the text, without increasing the size of the text
area.


        Stefan

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: rgm@gnu.org,  10960@debbugs.gnu.org,  l26wang@gmail.com
> Date: Sat, 10 Mar 2012 09:38:57 -0500
> 
> > Ping!  Stefan, could you please explain your response?
> 
> I'm not sure what to explain about it.
> 
> >> I don't understand Stefan's response in that bug: what effect exactly
> >> are text-scale-* functions supposed to produce?  Should the text in
> 
> Increase the size of the text, without increasing the size of the text
> area.

OK, let me ask specific questions then:

 . should the font of the text shown in the margin area change to the
   same size as the font of the text shown in the text area?

 . if the answer to the previous question is YES, then should Emacs
   automatically make sure the margin areas adjust their pixel width
   to keep the same number of character cells as specified by the
   set-window-margins command?

bug#10960: linum-mode does not work properly when changing font size

[Juanma Barranquero]

On Sat, Mar 10, 2012 at 16:06, Eli Zaretskii <eliz@gnu.org> wrote:

> OK, let me ask specific questions then:

In the thread of bug#1255 Stefan already answered what amounts to YES, NO.

The problem with that answer is that it is maximally unhelpful. Every
package that wants to use the margin will have to be defensively coded
against text-scale-*.

    Juanma

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Juanma Barranquero <lekktu@gmail.com>
> Date: Sat, 10 Mar 2012 16:38:14 +0100
> Cc: Stefan Monnier <monnier@iro.umontreal.ca>, 10960@debbugs.gnu.org, l26wang@gmail.com
> 
> On Sat, Mar 10, 2012 at 16:06, Eli Zaretskii <eliz@gnu.org> wrote:
> 
> > OK, let me ask specific questions then:
> 
> In the thread of bug#1255 Stefan already answered what amounts to YES, NO.

Maybe I misunderstood, but that's not what I read there.

> The problem with that answer is that it is maximally unhelpful. Every
> package that wants to use the margin will have to be defensively coded
> against text-scale-*.

If every application that uses display margins will have to do that,
Emacs should do that automatically, IMO.  What would be the reasons
not to?

bug#10960: linum-mode does not work properly when changing font size

[Stefan Monnier]

>> OK, let me ask specific questions then:
> In the thread of bug#1255 Stefan already answered what amounts to YES, NO.

Indeed.

> The problem with that answer is that it is maximally unhelpful.  Every
> package that wants to use the margin will have to be defensively coded
> against text-scale-*.

You mean, that text-scale-* makes the problem more common significant?
Maybe so, yes.  But I don't think that just special-handling
text-scale-* would be a good solution.  We need some more general way to
adjust the size of the margin to the size of the images&text displayed
there, so as to handle non-default faces, proportional text, ...


        Stefan

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Eli Zaretskii <eliz@gnu.org>,  10960@debbugs.gnu.org,  l26wang@gmail.com
> Date: Sat, 10 Mar 2012 13:28:52 -0500
> 
> > The problem with that answer is that it is maximally unhelpful.  Every
> > package that wants to use the margin will have to be defensively coded
> > against text-scale-*.
> 
> You mean, that text-scale-* makes the problem more common significant?
> Maybe so, yes.  But I don't think that just special-handling
> text-scale-* would be a good solution.

Special-casing it was not what I had in mind.

> We need some more general way to adjust the size of the margin to
> the size of the images&text displayed there, so as to handle
> non-default faces, proportional text, ...

It all boils down to calculating the size of the margins in pixels.
The key to that is window_box_width, which uses
WINDOW_FRAME_COLUMN_WIDTH to convert columns into pixels, and
WINDOW_FRAME_COLUMN_WIDTH returns the "canonical" width of a column,
which is an average width of the default face's font.

So all it takes is to make that calculation aware of face remapping,
images displayed in the margin, etc.

bug#10960: linum-mode does not work properly when changing font size

[Stefan Monnier]

>> You mean, that text-scale-* makes the problem more common significant?
>> Maybe so, yes.  But I don't think that just special-handling
>> text-scale-* would be a good solution.
> Special-casing it was not what I had in mind.

Actually, what you propose does just that.

>> We need some more general way to adjust the size of the margin to
>> the size of the images&text displayed there, so as to handle
>> non-default faces, proportional text, ...

> It all boils down to calculating the size of the margins in pixels.
> The key to that is window_box_width, which uses
> WINDOW_FRAME_COLUMN_WIDTH to convert columns into pixels, and
> WINDOW_FRAME_COLUMN_WIDTH returns the "canonical" width of a column,
> which is an average width of the default face's font.
> So all it takes is to make that calculation aware of face remapping,
> images displayed in the margin, etc.

OK, so rather than special casing text-scale-* you suggest to special
case frame-remapping: same problem.  It still won't account for
non-default faces, proportional fonts, and other display features,


        Stefan

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: lekktu@gmail.com,  10960@debbugs.gnu.org,  l26wang@gmail.com
> Date: Sat, 10 Mar 2012 20:36:29 -0500
> 
> >> You mean, that text-scale-* makes the problem more common significant?
> >> Maybe so, yes.  But I don't think that just special-handling
> >> text-scale-* would be a good solution.
> > Special-casing it was not what I had in mind.
> 
> Actually, what you propose does just that.

Please explain where do you see any special-casing.  I'm beginning to
think I cannot understand written English or write clearly in it.

> > It all boils down to calculating the size of the margins in pixels.
> > The key to that is window_box_width, which uses
> > WINDOW_FRAME_COLUMN_WIDTH to convert columns into pixels, and
> > WINDOW_FRAME_COLUMN_WIDTH returns the "canonical" width of a column,
> > which is an average width of the default face's font.
> > So all it takes is to make that calculation aware of face remapping,
> > images displayed in the margin, etc.
> 
> OK, so rather than special casing text-scale-* you suggest to special
> case frame-remapping

Where and how did you manage to read something like that into what I
said?

bug#10960: linum-mode does not work properly when changing font size

[Stefan Monnier]

>> > It all boils down to calculating the size of the margins in pixels.
>> > The key to that is window_box_width, which uses
>> > WINDOW_FRAME_COLUMN_WIDTH to convert columns into pixels, and
>> > WINDOW_FRAME_COLUMN_WIDTH returns the "canonical" width of a column,
>> > which is an average width of the default face's font.
>> > So all it takes is to make that calculation aware of face remapping,
>> > images displayed in the margin, etc.
>> OK, so rather than special casing text-scale-* you suggest to special
>> case frame-remapping
> Where and how did you manage to read something like that into what I
> said?

I guess I did not understand what you suggest, then.
Could you explain some more?
The way I see it, set-window-margins can only set the margin's size
either to "no margin" or to a fixed size given as an integer.  So you'd
have to let the Elisp code somehow figure out the *display* size of the
margin's text (either in pixels or in char-cell-equivalent), and I don't
know of any way Elisp can do that currently.


        Stefan

bug#10960: linum-mode does not work properly when changing font size

[Juanma Barranquero]

On Sat, Mar 10, 2012 at 19:28, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> We need some more general way to
> adjust the size of the margin to the size of the images&text displayed
> there, so as to handle non-default faces, proportional text, ...

OK, but that's a long way from your earlier answer in bug#1255: "So,
yes, the problem lies somewhat in linum-mode which should resize the
margin accordingly, tho it's far from easy for it to do so (and it can
only do it in increments of the base default font size)." We need a
way to adjust the size of the margin, but linum-mode et al. should do
nothing more** that set a margin width (in character cells); any
subsequent width change should be transparent for it/them.

**In fact, I don't see a way for two packages to coordinate uses of
the margin (let's suppose I write a package which uses the margin to
mark bookmarked places and I also want to use linum in the same
buffer); but that's unrelated to margin auto-resizing.

    Juanma

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: lekktu@gmail.com,  10960@debbugs.gnu.org,  l26wang@gmail.com
> Date: Sun, 11 Mar 2012 10:37:45 -0400
> 
> The way I see it, set-window-margins can only set the margin's size
> either to "no margin" or to a fixed size given as an integer.

That's correct, but all set-window-margins does is set two attributes
of the window object.  How to use that is entirely up to the display
engine.

When the time comes to display the window, the margins come into play
through the pixel coordinates where we draw the fringe bitmap.  The
code which computes those coordinates eventually calls
window_box_left_offset or window_box_right_offset, which call
window_box_width.  The latter is the single place in the code, AFAICS,
that converts the margin width from character cell units to pixel
units.

> So you'd have to let the Elisp code somehow figure out the *display*
> size of the margin's text (either in pixels or in
> char-cell-equivalent), and I don't know of any way Elisp can do that
> currently.

I didn't suggest to do this in Lisp.  (I don't think Lisp programs
should control the display at this low level.)  What I suggested is to
modify the calculation of the fringe pixel position based on how many
pixels are actually needed for the glyphs (character glyphs, stretch
glyphs, image glyphs, etc.) displayed in the fringe.  Isn't that what
you wanted?

bug#10960: linum-mode does not work properly when changing font size

[Stefan Monnier]

> I didn't suggest to do this in Lisp.  (I don't think Lisp programs
> should control the display at this low level.)  What I suggested is to
> modify the calculation of the fringe pixel position based on how many
> pixels are actually needed for the glyphs (character glyphs, stretch
> glyphs, image glyphs, etc.) displayed in the fringe.  Isn't that what
> you wanted?

If we can do that, it's great, yes.


        Stefan

bug#10960: linum-mode does not work properly when changing font size

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: lekktu@gmail.com,  10960@debbugs.gnu.org,  l26wang@gmail.com
> Date: Sun, 11 Mar 2012 21:54:09 -0400
> 
> > I didn't suggest to do this in Lisp.  (I don't think Lisp programs
> > should control the display at this low level.)  What I suggested is to
> > modify the calculation of the fringe pixel position based on how many
> > pixels are actually needed for the glyphs (character glyphs, stretch
> > glyphs, image glyphs, etc.) displayed in the fringe.  Isn't that what
> > you wanted?
> 
> If we can do that, it's great, yes.

I'll see what this would take when I have time.  I'm afraid we will
have to leave for post-24.1, though, as the changes will probably be
non-trivial enough.