I can confirm that Debian is also affected by this:
<http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=613190>. In all likelihood this just needs the test suite updating to reflect changes in the behaviour of Cairo, but it's not obvious to a non-Cairo user what this might be. Incidentally, the same bug has been reported at
<https://rt.cpan.org/Ticket/Display.html?id=62603>.

Created attachment 182536 [details] [review]
small patch in order to let the three test cases pass...

Hello,

It is possible to pass these 3 failing tests without disabling them, by
creating an extra surface that actually permits you to create
'color-alpha' surfaces with the `create_similar` sub.

At line 26 of t/CairoSurface.t, an 'rgb24' surface is created.
It seems (to me) that it is not possible anymore to create a
'color-alpha' surface from such an 'rgb24' surface with the
`create_similar` sub with the newest Cairo. I don't know if this
is intended behavior or not.

'argb32' surfaces seem to permit you to `create_similar`
surfaces in both 'color-alpha' or 'color' content.

The original test cases and these altered test cases also emit a
(non-fatal) warning: unknown surface type 16 encountered at
t/CairoSurface.t .. I don't know why (yet),

Perhaps this is useful,

Greetings,

Raymond.

Hello, are there any Cairo experts able to look at this issue, and perhaps review Raymond's patch?

Thanks,
Dominic.

The issue was actually caused by a premature call to $surf->finish.  Fixed in
stable-1-06 and master:
<http://git.gnome.org/browse/perl-Cairo/commit/?h=stable-1-06&id=9073da824de8>.