-
Notifications
You must be signed in to change notification settings - Fork 611
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Indic2 final-Reph reordering algorithm doesn't seem to match spec #2298
Comments
We started with the spec but then adjusted to match our understanding of what Uniscribe was doing. If you're suggesting a change we need to see tests and how Uniscribe handles them. Reading the code and comment it looks like the copying was very intentional. |
No, not suggesting a change. Really just wanting to understand the reason — for example, whether (a) it was discovered that step e in the spec produces incorrect results, or possibly (b) step e was a no-op somehow, or (c) it was discovered that Uniscribe actually didn't do step e, etc. It sounds like you're saying it was (c)? For the shaping doc repo, it's just important to record if there's a mistake in what the MS pages say. The usual caveats apply as to whether MS would change their page content if we reported an bug against it, but we can record the issue for future reference. |
I used
So yes possible that can be improved. I'm so far removed from that code right now, but maybe @jfkthame can suggest something. I don't have my extensive testing infrastructure set up. But maybe we do some testing setup in the cloud for realz this time. |
Cool. @adrianwong brought it to my attention in the docs repo, so it's possible he'd have testworthy material, too. Knowing it might be Kannada-related is a good start. (Super intriguing to me that nobody (AFAICT) has noticed any incorrect results from it over the years, so I wonder if it's only appearing in rare corner cases or if the spec doesn't actually need that step in it at all.) |
I read it some now. Looks like some of the step 5 we do in step 6. So it's just mispositioned with the comments:
followed by:
So step 6 says just move it to the end, but we move it to before A gentle reshuffling of code/comments without behavior change is what I suggest. |
Okay. Makes sense. Thanks! |
I will take a crack at finessing the comments if you like. |
Over in the opentype-shaping-documents repo, on issue 48 it was pointed out that the HarfBuzz final-reordering algorithm for Rephs in Indic2 seems pretty different from the MS script-shaping spec.
Most notably is adrianwong's comment that HB repeats step b at step e (that's not my analysis; the comment inline says the code was copied), when step e is very different in the Microsoft spec page.
Since this difference does not seem to have generated test errors for HarfBuzz, we were wondering what the rationale is for the step-b-to-e repeat. Or what else might be going on.
The text was updated successfully, but these errors were encountered: