Re: wxt terminal in windows hangs

[Bastian M. <bma...@we...>]

Am 18.01.2011 15:29, schrieb Allin Cottrell:
> On Mon, 17 Jan 2011, sfeam (Ethan Merritt) wrote:
>
>> On Monday, January 17, 2011, Allin Cottrell wrote:
>>> On Mon, 17 Jan 2011, Allin Cottrell wrote:
>>>
>>>> On Tue, 18 Jan 2011, Mojca Miklavec wrote:
>>>>>
>>>>> My bet is that fc-cache runs in the background to regenerate font
>>>>> index "every now and then", so that one can use any given font on the
>>>>> system.
>>>
>>> (and this can take a looong time).
>>>
>>> I notice that in gnuplot's cp_cairo.c there's a comment (around
>>> line 690) about "working around a bug in the cairo/win32 backend
>>> for font rendering": the work-around involves gp_cairo on Windows
>>> using the freetype + fontconfig backend by preference to the win32
>>> backend.  Following Mojca's lead, this could explain the initial
>>> long wait for gnuplot output.
>>>
>>> I'm wondering, is the cairo bug in question still there?
>>> Unfortunately the comment is not specific (no bugzille ID). But if
>>> the issue is fixed in current cairo it could be worth making a
>>> change. Cairo's win32 backend doesn't have the problem of a long
>>> wait on first use.
>>
>> Here's a link to the original thread reporting the problem
>> in cairo's win32 backend:
>>
>> http://groups.google.com/group/comp.graphics.apps.gnuplot/browse_thread/thread/5742a3d915085f7b/d075063d7da423e9?hl=en&ie=UTF-8&oe=utf-8&q=missing+pixels+for+Windows+PNG+driver&pli=1
>
> Thanks! I had a vague memory of that discussion but had forgotten
> the details. There have been a fair number of commits in the
> cairo-font-win32 area since then (14 months ago) so this is
> probably worth revisiting. I'll try some testing with cairo 1.10.2
> on Windows.
>
> Allin Cottrell

I have tried to reproduce the original bug report of the font problem
in cairopng with pango 1.28.3 and cairo 1.10.2 on Windows. The result
using current CVS (freetype + fontconfig) looks like this:
http://www.physi.uni-heidelberg.de/~bmaerk/tmp/gretl_bug.png
And here's the result using the win32 backend:
http://www.physi.uni-heidelberg.de/~bmaerk/tmp/gretl_bug-win32.png
Both versions look fine to me.

Should we apply the revised test you proposed?
 > #if CAIRO_VERSION_MAJOR < 2 && CAIRO_VERSION_MINOR < 10
Or should we test for the pango version here?

There's another reason why I would prefer that the cairo/pango
terminals use the win32 backend: In my experience, these
terminals (read: cairo/pango/fontconfig) fail to find a sensible default 
font ("Sans") on 4 out of 5 Windows machines. For me,
it fails on three different XP machines and on a Win 7 (x64)
machine. It only works on a Vista machine. This is the
result of 'enhancedtext.dem' using the pngcairo terminal with
no font specified:
http://www.physi.uni-heidelberg.de/~bmaerk/tmp/fontconfig-bug.png

   Bastian