Le 21/03/2019 à 00:08, Richard Kimberly Heck a écrit :
I can confirm this this helps a lot on windows, at least with the test
case that we received.
Riki, OK to backport?
Yes, thank you.
Is there any danger to this, do you think? If so, I could try to produce
a 'testing' binary before we move toward the real release.
I I did not mistype something, the danger is null: we already cache
rbearing (and width, ascent and descent) for the same reason. I only
added caching of lbearing. The only use of lbearing has been added in
the 2.3 development cycle, whence the regression on windows.
The patch looks more complicated than it really is because git diff is
trying to be too clever. If you look at the code after the patch, you
will see that the structure of lbearing is exactly the same as the one
of rbearing.
JMarc
