Re: [ft-devel] Problem with Postscript Hint

[John Tytgat]

On 5/3/2011 9:52 AM, PILLET Céline wrote:

I have a problem with hints of PostScript fonts.

I create 2 consecutives bitmap with a postscript font, with a multithread program, and rendering differs slightly.

I don’t have this problem when I create bitmap without “hints”.

I don’t have this problem when I create bitmap with “hints”, in a single thread program.

Do you have an idea of ​​something that is not thread safe in the PostScript hints code?

If you mean 'PostScript font' = Type 1, CID or CFF font', search for 'seed' used for implementing a random generator.  This is calculated based on location of stack variable.  This *could* explain what you're experiencing but I'm not sure.

Question for the FT developers : why was this seed calculation implemented this way ? And not e.g. a fixed start value associated with each FT_Face created ? That would make FT output reproducable (and testable for detecting regressions).

John.