Subject: Re: kern/35739: recent addition of RASOPS_ROTATION breaks wsfont
To: None <email@example.com>
From: Miod Vallat <firstname.lastname@example.org>
Date: 02/21/2007 22:29:03
> The RASOPS_ROTATION stuff breaks the wsfont abstraction, since callers now
> mess with the font data on their own. If I recall, this means fonts in use
> by other displays can have their glyphs changed.
No, they don't. Rotation is done in wsfont and in a copy of the original
font data. Rasops maintains a list of (normal, rotated) font cookies
tuples, in order to avoid requesting the same rotated font from wsfont
> Code inspection
You might want to read it again.
> rasops shouldn't mess with the font data. The 'rotation' should be another
> attribute encoded in the cookie returned by wsfont_find(), and if the font
> needs to be rotated it should be done by wsfont_lock(), not the caller. Once
> that's done, the rotation code in wsfont shouldn't be conditionalized - it's
> small enough.
This looks indeed like a good way to fold this code in the existing