A number of the example-illustration images in the Emoji shaping document are misaligned.
The first batch of those was created quite some time ago using the same method as all the other illustrations:
Generate isolated glyph images atomically with hb-view
Concatenate them together with the montage command from Imagemagick.
For reasons I haven't been able to figure out, that approach broke when I regenerated a bunch of images back in June while updating the whole document; I'm not sure if there was an API change in an Imagemagick update that I overlooked or if hb-view was giving slightly different results because I had to pass it different flags to handle some of the emoji distinctions. I poked around a tiny bit but couldn't figure it out, and have been swamped enough with school work that I couldn't afford to delay the merge any longer.
It does seem like the tileing behavior for Imagemagick has changed somewhat from before.... The emoji sequences are quite long, but so are several of the sequences used by illustrations in many of the complex-script shapers, and I didn't run into unexpected -tile hurdles there....
So at some point I hope I can figure out what the issue is, but in the meantime anyone is welcome to read the image-generation log and see if they can reproducibly get better results.
The log files are not quite automatable scripts (yet?); that was the original intent of them but I haven't gotten to that yet. Individual commands should work, but you may have to adjust paths for the font binaries (and in the Emoji doc, there are quite a few of those, given the complexity of emoji engineering). Please comment if you are poking at that and have trouble locating a particular binary, of course.
A number of the example-illustration images in the Emoji shaping document are misaligned.
The first batch of those was created quite some time ago using the same method as all the other illustrations:
hb-viewmontagecommand from Imagemagick.For reasons I haven't been able to figure out, that approach broke when I regenerated a bunch of images back in June while updating the whole document; I'm not sure if there was an API change in an Imagemagick update that I overlooked or if hb-view was giving slightly different results because I had to pass it different flags to handle some of the emoji distinctions. I poked around a tiny bit but couldn't figure it out, and have been swamped enough with school work that I couldn't afford to delay the merge any longer.
It does seem like the
tileing behavior for Imagemagick has changed somewhat from before.... The emoji sequences are quite long, but so are several of the sequences used by illustrations in many of the complex-script shapers, and I didn't run into unexpected-tilehurdles there....So at some point I hope I can figure out what the issue is, but in the meantime anyone is welcome to read the image-generation log and see if they can reproducibly get better results.
The log files are not quite automatable scripts (yet?); that was the original intent of them but I haven't gotten to that yet. Individual commands should work, but you may have to adjust paths for the font binaries (and in the Emoji doc, there are quite a few of those, given the complexity of emoji engineering). Please comment if you are poking at that and have trouble locating a particular binary, of course.