Bug 175778 – Perform special handling for whitespace and control characters

• Component: Text
• Status: NEW

[Myles C. Maxfield]

Currently, we treat these characters just like any other characters, which means we perform font selection, and possibly render the .notdef glyph. Instead, we should educate WebKit about these characters so that we don't need fonts for them (for performance) and to guarantee they won't render.

A few options:
U+200B ZERO WIDTH SPACE
U+FEFF
unicode bidi characters
U+2060 WORD JOINER
& more

[Myles C. Maxfield]

The fix for this would be platform-non-specific.

[Radar WebKit Bug Importer]

<rdar://problem/33996276>

[Dan Jacobson]

Please do not render U+FEFF ZERO WIDTH NO-BREAK SPACE

$ /usr/lib/x86_64-linux-gnu/webkit2gtk-4.0/MiniBrowser https://wiki.openstreetmap.org/wiki/List_of_OSM-based_services

It looks terrible!

[Dan Jacobson]

See also Bug 175489.
I'm worried people will complain to webmasters.
They probably have never seen such usually invisible characters now suddenly showing up in all their browsing years.

[Dan Jacobson]

This should be almost as high priority a bug as if the browser inserted
U+002E FULL STOP where it was not supposed to be, e.g., turning
12345 into 123.45!

[Dan Jacobson]

Created attachment 330323 [details]
Inspect Element in Epiphany

[Myles C. Maxfield]

This bug is about not rendering these special characters. Currently, we treat these characters the same way as any other character, so we will look up fonts that can render them, ask those fonts for glyphs for those characters, and render whatever the font gives us. Instead, we should have explicit code in WebKit that specifically checks for these code points and simply ignores them (so they have no glyph and no advance width).

[Myles C. Maxfield]

*** Bug 185976 has been marked as a duplicate of this bug. ***

[Myles C. Maxfield]

In https://bugs.webkit.org/show_bug.cgi?id=185976 I naively deleted most of the overwriteCodePoint() calls inside createAndFillGlyphPage(). Simply deleting those is probably incomplete; what we probably should be doing is having deliberate handling for these characters, rather than swapping the characters with other characters and treating them normally (performing fallback, etc.)