broken glyphs in github

Stream: general

[Mario Carneiro]

Is anyone else seeing missing glyphs in the definition of src#PointedCone ?

[Martin Dvořák]

Mario Carneiro said:

Is anyone else seeing missing glyphs in the definition of src#PointedCone ?

Can you post a screenshot?

[Mario Carneiro]

image.png

[Mario Carneiro]

it appears correctly for a moment but then gets messed up by the syntax highlighting or something

[Martin Dvořák]

On my computer, it is all right.
image.png
My font is Jetbrains Mono.

[Mario Carneiro]

this is in github

[Mario Carneiro]

it's fine locally

[Martin Dvořák]

Ah, my bad! I see black background and immediately thing it is VS Code.

[Mario Carneiro]

Here's what it looks like if I screenshot the page just after initial load
image.png

[Notification Bot]

9 messages were moved here from #PR reviews > #7386 Linear Programming by Mario Carneiro.

[Floris van Doorn]

For me it looks fine on Github, even when waiting a bit:
image.png

[Mario Carneiro]

(FF + ubuntu)

[Floris van Doorn]

Ah, it works for me in Windows + Chrome, but in Windows + Firefox it is also messed up:
image.png

[Patrick Massot]

Same here.

[Anatole Dedecker]

Yes I've been having these issues in Firefox (Ubuntu) for a while :/

[Eric Rodriguez]

FF+OS X has this issue too (as does FF+Android, but in a more serious way that has been known for a while)

[Kevin Buzzard]

(indeed it's also bad for me in FF, fine in Chrome)

[Luigi Massacci]

Surprisingly it works just fine with Firefox + Ipad OS, unexpected

[Ruben Van de Velde]

For now, Firefox on iOS is actually safari under the hood (though that may change in Europe)

[Mario Carneiro]

Those look like unpaired surrogates in UTF-16, I'm guessing there is some bad unicode handling in github's code

[Kevin Buzzard]

Didn't @Reid Barton have an example of this years ago? Or is this a different issue?

[Eric Wieser]

We saw a version of this before where the syntax highlighting would cut a surrogate pair in two

[Mario Carneiro]

It looks like the code has been split into two-UTF-16-codepoint chunks, which is why only some of the 𝕜's are broken
image.png

[Eric Wieser]

(the previous bug was declared wont-fix due to lack of developer time)

[Mario Carneiro]

do you have a link @Eric Wieser ? This does indeed seem to be the same bug

[Kevin Buzzard]

I'm pretty sure there was an elegant one-line gist with some \< s in.

[Eric Wieser]

https://github.com/github-linguist/linguist/issues/5069

[Mario Carneiro]

Actually that one seems to be a different issue, that's about splitting across different color regions. Here there is splitting within a single chunk, surely there isn't anything in the regex highlighter saying you should split the text into two char chunks

[Mario Carneiro]

also the original reproduction is no longer broken on github

[Kevin Buzzard]

Well that issue was certainly what I was referring to

[Mario Carneiro]

In chromium the span is not split into pieces:
image.png

[Mario Carneiro]

so this seems like an issue with github proper and not linguist

[Mario Carneiro]

and/or firefox

[Mario Carneiro]

reported as https://github.com/orgs/community/discussions/103098

[Kevin Buzzard]

I gave it an upvote :-)

[Floris van Doorn]

This is now fixed in Firefox 124+.