Re: GtkPrintOperation uses wrong scale on Windows ?

[Geert Janssens <info kobaltwit be>]

On 27-06-12 11:34, Geert Janssens wrote:
> On 26-06-12 16:29, Michael Cronenworth wrote:
> > Geert Janssens wrote:
> > > Am I doing something wrong here ?
> > Was GNUCash setting units before? If so, what to?
> It was. This setting has never changed since 2007, while the bug first 
> appeared in 2010. Your question make me realize I may be chasing the 
> wrong tail here.
>
> Instead I am now comparing changes between the last known working 
> version and the one first reported to be failing. There's only about a 
> hundred commits in between and each rebuild on my Windows test machine 
> takes a couple of hours... :(
>
> I have identified one change that could be a good candidate and I'm 
> now testing my theories with this one reverted. We'll see the result 
> later today.
>
> Thanks for the feedback so far.
>
> Geert
> _______________________________________________
> gtk-app-devel-list mailing list
> gtk-app-devel-list gnome org
> https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list

I found my issue, and it was indeed a completely different one. I'll 
share my experience here in case others hit the same problem.

The GnuCash code called cairo_identity_matrix on the cairo_t obtained 
via gtk_print_context_get_cairo_context with the intention to start form 
a clean cairo_t to write our checks to.

cairo_identity_matrix removes all transformations from a cairo_t, in the 
process resetting its resolution to 72dpi. On linux this doesn't make a 
difference because GtkPrintContext works with a 72dpi resolution internally.

On Windows though, the cairo_t returned by 
gtk_print_context_get_cairo_context has the resolution of the printer 
(eg 600dpi). I suppose GtkPrintContext already set a transformation on 
the context to achieve this. Resetting the cairo_t also removes this 
internally set up transformation, ruining the complete page setup.

I don't think this description is not 100% accurate. I believe a cairo_t 
doesn't really have a resolution. But in the context of printing this is 
close enough to explain what went wrong.

I rewrote my code to no longer use cairo_identity_matrix (altering some 
of our own transformations to cope with this), and now the pages print 
fine on Windows as well.

Geert