1731723 - Missing glyphs in PDF rendering

Status: RESOLVED DUPLICATE of bug 1734853

(Core :: Graphics: Text, defect)

Description

[Martin Thomson [:mt:]]

I'm reading a lot of crypto papers lately. Depending on the zoom level, different glyphs are not being rendered, especially in mathematical formulae. It is not strictly the case that zooming in causes details to be restored; some details disappear at higher zoom. At some point zooming in does seem to resolve the issue, but it is not always the case that a higher zoom makes more details visible.

Take page 5 of https://fc13.ifca.ai/proc/5-1.pdf as an example. At 100% zoom, the "t ←r −cx mod q" in Figure 1 does not render the subtraction symbol. At 90% or 110% zoom, this symbol reappears. The formula in the next section, "yj := gxj (j ∈ [n])" does not render the ":=" until the zoom level hits 170%. Before that it variously hides the ":" and "=", sometimes rendering the "=" as something like an en-dash.

Take page 5 of https://petsymposium.org/2018/files/papers/issue3/popets-2018-0026.pdf Section 3.2, item 2. Again, "s = (t −ck) mod q" at 100% renders as "s - (t −ck) mod q". At 110% the "=" disappears entirely. At 130% (closer to what automatic zoom chooses for me), this renders as "s = (t ck) mod q", losing the subtraction sign.

I can understand that this might be a font quality issue, but that does not entirely absolve the renderer of responsibility. These papers seem to be free of these rendering issues at 80% zoom. Of course, that zoom level makes the text nigh unreadable on my screen.

(Happy to provide a few screenshots if others are unable to reproduce.)

---

Compositing WebRender (Software D3D11)
WebGL . Driver Renderer Google Inc. (NVIDIA) -- ANGLE (NVIDIA, NVIDIA GeForce GTX 1080 Direct3D11 vs_5_0 ps_5_0, D3D11-30.0.14.7212)
Direct2D true

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:jfkthame, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 2

[Jonathan Kew [:jfkthame]]

This looks like an issue at the rasterization level, not layout; moving to Graphics.

I'm not seeing this on my Windows machine, but it's may well be dependent on details of the graphics system and/or font rendering settings. Do you have ClearType enabled? What resolution (system display scaling factor) does your system use?

Screenshots would be interesting, to see how the fonts are rendering for you. I do see some unevenness -- e.g. the minus sign looking extremely thin, though still visible, or the two strokes of the equals sign being different weights -- which suggests a lack of hinting (or poor hinting) in the font involved.

Comment 3

[Martin Thomson [:mt:]]

Yes, ClearType is/was on. The monitor is at 100% scaling and I'm using the native resolution of the monitor.

...and now that I try to get screenshots, it's no longer a problem. Just the same unevenness that Jonathan mentions, which is a common problem for LaTeX fonts. I've verified that there was no change in graphics drivers or hardware, just the usual Firefox Nightly updates.

Comment 4

[Martin Thomson [:mt:]]

I just downloaded an old Firefox build (around the time I noticed the issue) and was not able to reproduce there either.

Comment 5

[Jonathan Kew [:jfkthame]]

I think bug 1734853 may be another report of the same issue; might be clues there to help with reproducing.

Comment 6

[Martin Thomson [:mt:]]

Attachment: 100%

Ah, that did it. Setting gfx.webrender.software to true and restarting the browser caught it. Screenshots!

Comment 7

[Martin Thomson [:mt:]]

Attachment: 90%

Comment 8

[Martin Thomson [:mt:]]

Attachment: 110%

This one doesn't seem to have any (edit: many) issues today ¯\(ツ)/¯

Comment 9

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:lsalzman, could you have a look please?

For more information, please visit auto_nag documentation.