[ft] Glyph's Bitmap Beyond Ascender

[Mickey Gabel]

Hello, again.

I've encountered a strange problem, where it appears as if a glyph is 
rendered beyond the maximum ascender, as well as the maximum bbox as 
reported by the FaceRec. I understand that ascender is not always 
reliable, but as I understand from the docs, the bbox IS reliable.

I'll try to get a small reproducible code, but it will probably take 
some time, for now here is some information:

* FreeType 2.3.5.
* I don't use the patented hinter (ft_option.h is the default one).
* Font is "JI-acinic.ttf" (which probably came with some software I 
installed).
* I haven't set any transformation, and all my own transformation use 
whole pixels, no fractionals. I also make sure to round, ceil or floor 
all vectors from FT2 (even though it seems those I get are rounded anyway).
* The I use Set_Char_Size with 32pt x 32pt, resolution 96x96 dpi.
* Antialiasing is on, so is hinting.
* FreeType reports ascender of 35. Going by face->bbox we get yMax of 
813, which makes sense (1000 units per em, 43 pixels per em, 813 * 43 / 
1000 gives 34.959).
* When rendering lowercase 'e', I give FT_Glyph_To_Bitmap() an origin 
vector with y = 9 * 64. There are no other transformations or 
translations using FT.

Therefore I expect that bitmap_glyph->top will be 44 or smaller (9 + 35 
is 44). However, I get top of 45, which means that the 'e' character 
rises beyond its ascender AND beyond the bbox (converting the yMax to 
font units, we get 1046 == 45 * 1000 / 43, which is beyond the 813 
reported by the FaceRec).

Turning off hinting using FT_LOAD_NO_HINTING does help, but the ascender 
is already rounded up (34.969 --> 35), as suggested in the Glyph 
Convention doc. Could hinting move it up even more?

I do need a reliable way to figure out the maximum character height for 
a face before loading glyphs (to allocate images, for example, and for 
layout).

Am I doing something wrong? Is this a bug? Any ideas?

Thanks,
        Mickey