Re: Trunk emacs infelicity with linum mode

[martin rudalics]

>>   > If I set a breakpoint in Fface_font, which is what barfed in the OP's
>>   > case, the breakpoint doesn't break inside the call to
>>   > adjust_frame_size or Fx_create_frame, it breaks after Fx_create_frame
>>   > already returned.  I don't know why the difference.
>>
>> I suppose the OP confused two calls of adjust_frame_size - the one at
>> line 4609 and the one at line 4678.
>
> Or maybe we confused that.

We won't be able to tell but among the values he posted we have an
old_text_width of 10 which is inherently incompatible with a call from
line 4678 and a frame column width of 9.

> Probably because of the tool bar or the scroll bars -- these are the
> only differences between the builds, AFAIR.

The tool bar is set before calling x_figure_window_size so I suspect
it's the scroll bar height call which in adjust_frame_size already sets
all values just as the change_frame_size callback triggered by the call
on line 4678 wants them.  This would explain that the hook is _not_
triggered by the call on line 4678.  It doesn't explain why it's
triggered on Motif/Lucid where I basically "do the same thing" (though
apparently I don't).

> What will break if we set f->official true only _after_ the second
> adjust_frame_size returns?

Nothing, I presume.  More so, because the fact that the hook is not
triggered on Windows and Gtk, already constitutes a bug which makes the
use of the hook during frame setup impractical on these systems.  Note,
however, that the hook might still trigger any time in between setting
f->official and executing

(face-set-after-frame-default frame parameters)

though currently I can't see how that could happen.

martin