Jason Rumney wrote:
> The new code seems to be making a poor choice of font for Latin-1 
> characters, it is choosing a symbol font which does not contain the 
> characters for me. This was triggering a bug that is also present in 
> the trunk which caused the crash that Juanma experienced. The choice 
> of font for other Latin characters is also not optimal, as the font 
> used for ASCII contains most of these characters, but in the 
> font-backend branch other fonts are chosen.

Here is an example of how display in the font-backend branch differs 
from the trunk. I have highlighted some areas in the two images that 
look interesting for figuring out what might be going wrong: