Re: pango_cairo_update_layout()

[Theo Veenker <Theo Veenker let uu nl>]

Behdad Esfahbod wrote:
> Thanks Theo.  I can reproduce this.  Will look into it as time permits.  In
> the mean time, mind filing a bug on bugzilla.gnome.org against pango so it
> doesn't get lost?
>
> behdad

I will do so. Thank you.
Theo


> Theo Veenker wrote:
>  > Behdad Esfahbod wrote:
>  >> Theo Veenker wrote:
>  >>> Hi,
>  >>
>  >> Hi,
>  >>
>  >>> A question regarding pango_cairo_update_layout(). The example in
>  >>> http://library.gnome.org/devel/pango/stable/pango-Cairo-Rendering.html
>  >>> contains the following fragment:
>  > [snip]
>  >>> Why is it that it doesn't make a difference whether
>  >>> pango_cairo_update_layout()
>  >>> is called or not. The output is the same. Is it really necessary to 
> call
>  >>> this
>  >>> function if only the cairo context changes? Why? In my own code I also
>  >>> don't
>  >>> notice any diffrence if I just drop the pango_cairo_update_layout()
>  >>> call.
>  >>
>  >> It's very delicate.  For example, if you have hinting enabled, then
>  >> the size
>  >> of a layout with identity matrix may be different from one of a
>  >> rotated matrix
>  >> as hinting is only done with axis-aligned transforms.  And for pango
>  >> to pick
>  >> up the matrix from your cairo_t you need to make that function call.
>  >
>  > Thanks for the info. But I now noticed some condition where calling
>  > pango_cairo_update_layout() has a big unexpected impact. I've created
>  > a little X test application that demonstrates it (attached). The program
>  > draws a string using cairo-pango. It responds to following keys:
>  >  - page up/down: controls scaling factor for cairo_scale()
>  >  - up/down: controls rotate angle for cairo_rotate()
>  >  - 1/0: enable/disable calling pango_cairo_update_layout() before
>  >         pango_cairo_show_layout()
>  >  - r: recreate layout object
>  >  - R: recreate both context and layout object
>  >  - enter: dump image as xresult.png
>  >
>  > Start the program and scale up to a factor 5 (hit pgup). It now looks
>  > like in attached result1.png. Hit '1' to enable calling
>  > pango_cairo_update_layout().
>  > Now you'll see result2.png, which is rather weird. Turn off calling
>  > pango_cairo_update_layout() (hit '0') and recreate the layout (hit 'r').
>  > The result stays bad. After recreating both the context and layout (hit
>  > 'R')
>  > the output is OK again. FWIW, there is no difference (in my eyes) 
> when the
>  > image is rotated a bit.
>  >
>  > What's going on here? I can't believe it is a font question as the 
> output
>  > looks perfect to me when not calling pango_cairo_update_layout(). 
> Hopefully
>  > it is just a silly error on my side.
>  >
>  > Could you please take a look at this Behdad?
>  >
>  > Regards,
>  > Theo
>  >
>  >
>  > ------------------------------------------------------------------------
>  >
>  >
>  > ------------------------------------------------------------------------
>  >