Oleg Oshmyan

To the best of my understanding, for this branch to be taken: ```c if (DIFF(l1_new, l2_new) < DIFF(l1, l2)) { if (w->linebreak || w == text_info->glyphs) ``` it must be...

> From a quick check in Aegisub it seems like a single trailing `MUSTBREAK` does not result in vertical space being added, this only happens for _multiple_ trailing `\N`s? Yes,...

Why _does_ the line-break-after-the-last-character matter for our line breaker anyway? Isn’t that something we’d want to fix?

Sorry for making you repeat that. Am I inferring correctly that the forced-linebreak branch forces zero advance width for that glyph but we still render the glyph itself? (As in...

> The forced break branch never looks at the length of the current line or glyph; it just sets the next character as the begining of a (forced) new line....

> If not bidi, won't contextual shaping be influenced by the presence or absence of linebreaks? Line breaks themselves, yes. But if we wanted to hide the glyphs _within_ the...

I believe MrSmile has wanted to work on this for a long time. To me though, it’s not obvious how to do this: you need to account for blur of...

I by no means want to discredit the importance of offscreen clipping, but: > animations that result in large amounts of text offscreen are liable to be cache-miss-heavy anyway (like,...

Hang on. The text has a 3D transform and a script resolution of 1280×720, but the video resolution is 1920×1080. This means the transform is distorted. Testing this in VSFilters...

I mean, you’re not wrong. The actual cause of the slowdown is indeed our lack of offscreen clipping. I just wanted to point out that VSFilters other than MPC-HC suffer...