Tk Source Code

View Ticket
Login
Ticket UUID: 4be479d0855979ad8718cdff7fed54a9223aabc
Title: tkUnixRFont.c: nothing defines XFT_HAS_FIXED_ROTATED_PLACEMENT
Type: Bug Version: 8.6.13
Submitter: chrstphrchvz Created on: 2023-11-23 22:02:57
Subsystem: 46. Unix Fonts Assigned To: nobody
Priority: 5 Medium Severity: Minor
Status: Open Last Modified: 2026-11-07 13:19:11
Resolution: None Closed By: nobody
    Closed on:
Description:

Searching for this macro online only yields its one use in TkDrawAngledChars(), which has been there since the patch in [1611359] proposing the TIP 119 implementation. I have not found any discussion of this macro. It has never been defined by Tk, Xft, or anything else. So I suspect the code defined by this macro has been unused since it was submitted.

Yet the code has had some updates made to it specifically, such as fixes/comments for [d9fdfa435d] and [71b1319acc]. Were those changes aware of this macro and actually tested with it defined?

It looks like the code was waiting on Xft to make some particular change. Did that ever happen? Or should the code just be removed?

User Comments: chrstphrchvz added on 2026-11-07 13:19:11:

Not asking for something to be done about this ticket anytime soon. Just pointing out the comment above TkDrawAngledChars(), which I finally bothered to read and probably explains the macro:

This would be simple code, except Xft has bugs with cumulative errors in character positioning which are caused by trying to perform all calculations internally with integers. So we have to do the work ourselves with floating-point math.

I have not tried defining XFT_HAS_FIXED_ROTATED_PLACEMENT and seeing what happens. But briefly glancing at e.g. XftGlyphExtents() in recent X.Org libXft, I doubt that the "bugs" referred to by the comment were ever fixed. And maybe the use of structures with integer coordinates (e.g. XGlyphInfo) in public APIs makes them difficult or impossible to fix.


fvogel added on 2023-11-25 09:00:35:

[a2e77573] that has introduced the XFT_HAS_FIXED_ROTATED_PLACEMENT #define was committed by dkf as an implementation by Simon Geard of TIP #119. Maybe one of these people can remember (although this happened 15 years ago...)?

Regarding the changes within that code, they originally came from the revised text implementation, see Gregor's bug report here as mentioned in the ticket you found [d9fdfa435d] that as a fix backports Gregor's fix from the revised_text branch. I could never observe the problem that Gregor's patch was supposed to fix, but I realize now that it might have been due to XFT_HAS_FIXED_ROTATED_PLACEMENT being undefined in my setup. Perhaps.

The other changes you mention from ticket [71b1319acc] would be better discussed with their author, Christian Werner (chw).


chw added on 2023-11-24 11:38:23:
Howdy Christopher,

last year or so I sent a big patch to the libXft repo which fixed
issues in applying transformations affecting e.g. rendering of color
emoji bitmap fonts, but fixing shear on regular fonts to get an italic
look, too. This is now part of the latest libXft source code. No idea
however, how far and/or if at all this has positive consequences for
using this macro. And since old libXft versions are around, how it can
be solved regarding versioning.