Place mark feature before mkmk. Fixes #377 #379
Conversation
|
@simoncozens @ohbendy : I'm not sure the marks are placed properly now here's the sample with the newly built font (I haven't published it yet). Ben's old submission build (when he sent us the sources we got also binaries) is reddish and the newly updated source font is greenish (see the mark positioning -- IMHO, it appears to be off by 1, but I might be wrong). Tested it for NotoLoopedLao but all of the Looped fonts have the same property.
|
|
Ok. I’m happy to look at this some more, but it highlights the need for the kind of shaping regression test framework that I recently added to fontbakery. (See https://simoncozens.github.io/tdd-for-otl/ ) Could you split off these sources to their own repos? That way I can add specific test suites for them. |
|
Yep unfortunately that's broken in a different way, glad you didn't publish yet! Agree there should be a testing stage. IMO the best person to do the checking would be the designers, because different people will choose to handle things in different ways, they will know the exact checks to run on each set of fonts, and will know what the desired result is. We can also check in different environments (e.g InDesign, Word, browsers) as unexpected things can occur. |
|
I'll create notofonts/Noto-Lao and notofonts/Noto-Thai. |
|
That is correct. Marek advised that the minority languages won't appear in a UI environment. |
|
OK. That makes sense. Some lookups were still there. |
|
Sorry for not removing those lookups, thanks for checking. Not sure what you've merged; the document fonts and UI fonts have different outlines and behaviour. |
|
By merging I mean put the masters of NotoLoopedLaoUI in NotoLoopedLao with an axis for the UI variation. The behaviour differences are:
For the master specific mark, an alternative solution would be to use a contextual composite glyph with the expected kerning. With the merged sources, lookup LaoContexualKerning would only work with the feature syntax update for font variation Simon has worked on, and the contextual composite glyph would work with current feature syntax. Otherwise, the fix for the mark and mkmk features is to have the kern feature before the mark and mkmk feature. I've already done most of this in the ui-as-axis branches on https://github.com/notofonts/Noto-Thai and https://github.com/notofonts/Noto-Lao. |
|
Ok. Please let me test the compiled fonts before publishing. |
|
The point of splitting to a separate repo is that you and I can design a test suite together, and then we can gate publishing on that test suite passing. |
The document and UI fonts also have different values for family alignment zones, WinAscent/WinDescent, underline and strikethrough thickness and position.
I just checked all the sources I'd sent and couldn't find anything that shouldn't have been there. Glyphs would have generated an error if there were lookups for nonexistent glyphs.
Sorry, no. Please just let me test the fonts before publishing. I have all the testing documents I need already, they're not part of the deliverables we agreed. This work was completed in December. I can't work for free on designing more tools for testing. |
Understood. At the very least, I want to add regression tests to ensure that the original bug you reported (#377) are fixed and that whatever is going on in Marek's example above is also fixed, and to ensure that these problems don't recur in future. But I can write those myself. |
See discussion in #377.