On 01/03/2025 21:28, Pip Cet wrote:

> Just a heads-up: if uniscribe would benefit from a general mechanism to
> avoid calling the font_close method twice, it may well be worth doing
> it; the other font backends have been fixed to make font_close
> idempotent, but people are unlikely to stop writing font backends
> anytime soon :-)
> 
> Thanks for working on this!  If there's a uniscribe-specific fix that
> works, feel free to disregard this email.


The fix is simple, just setting dwrite_cache = NULL, as you already 
worked out.

 From what I see, w32font_close is already idempotent, as it tests for a 
non-null hfont.

Thanks, and my apologies for not attending to this before.