Open Bug 1658558 Opened 1 year ago Updated 2 months ago

[ARM] Pages with various formats are not printed completely (several letters are arbitrarily missing)

Categories

(Core :: Printing: Output, defect, P2)

ARM64
Windows 10
defect

Tracking

()

Tracking Status
firefox79 --- affected
firefox80 --- affected
firefox81 --- affected

People

(Reporter: cornel_ionce, Unassigned)

Details

(Whiteboard: [print2020])

Attachments

(2 files)

Attached image printing_output.JPEG

Affected versions

*81.0a1 (BuildId:20200811091731)
*80.0b6 (BuildId:20200807195315)
*79.0 (BuildId:20200720193547)

Affected platforms

  • Windows 10 64-bit (arm64)

Steps to reproduce

  1. Launch Firefox.
  2. Go to https://drive.google.com/file/d/0BzOV3ybvBASXX3EyYzNMWllqVG8/view - download the file and open it in Firefox.
  3. Go to Menu -> Print.
  4. Select Landscape Mode and print the page.
  5. Compare the actual print with the print preview.

Expected result

  • The article is printed without any issues.

Actual result

  • Several letters are missing from the print version (see attached).

Additional notes

  • This issue is not a regression, it reproduces back to 2019-02-08 Nightly.
  • It also reproduces on Portrait mode.
  • The printer used was a HP LaserJet Pro MFP M28a.

Suggested severity

  • Since Printing is an important feature of Firefox but this issue only affects ARM architecture, I think this should be an S2.
Has Regression Range: --- → irrelevant
Has STR: --- → yes

jfkthame, does this ring any bells? I feel like we've seen this sort of issue before.

Severity: -- → S2
Whiteboard: [print2020]
Flags: needinfo?(jfkthame)

Cornel, does this happen 100% of the time for you?

And when it does happen, is it always the same letters that are missing?

Flags: needinfo?(cornel.ionce)
Summary: [ARM] Pages with various formats are not printed completely → [ARM] Pages with various formats are not printed completely (several letters are arbitrarily missing)
Attached file testcase 1

(here's the testcase that's linked in the initial comment here, BTW. I would expect you can load this bugzilla-hosted version directly, and the bug would still reproduce, which saves the "download & open it locally" steps.)

In the photo, the specific letters that seem to be missing are:

  • Lowercase letter "s" and "f", in the italicized version of the default font.
  • Lowercase letter "p" and "b", in the default font ("p" is used & missing in text of various sizes, too -- the default size, and 10pt, and 18px. "b" is only used [& missing] at the default size.)
  • Capital "N" in Times New Roman font.

This looks reminiscent of issues we see on 32-bit Windows 7 that we believe are due to the OS graphics system coming under too much pressure (resource constraints, maybe?) and directwrite returning errors when trying to rasterize some glyphs. I'm surprised to see it happen on a 64-bit win10 system, though. Perhaps there's no relationship to the 32-bit/win7 flakiness.

Would be interesting to know how consistent/reproducible it is. Always the exact same fonts/glyphs affected, or is it erratic?

Flags: needinfo?(jfkthame)

(In reply to Daniel Holbert [:dholbert] from comment #2)

Cornel, does this happen 100% of the time for you?

And when it does happen, is it always the same letters that are missing?

Yes, it happens all the time (printed 10 copies) and it seems that the same letters are missing.

Flags: needinfo?(cornel.ionce)
Priority: -- → P2

Sorry this has stalled out -- does this still happen? A lot of print changes have happened over the next year.

[Also, I'm bumping this to S3 to reflect reality, since we've been shipping with this for years without dupes being filed, and it's arm64-specific which means not a ton of users are affected, I think. Would be good to make progress here, but I'm not sure any of the relevant engineers have arm64 hardware handy... we'd probably need to ship out some machines.]

Severity: S2 → S3

FWIW, I'd find it a bit surprising if this is truly dependent on arm64 vs intel architectures; my hunch is that it could be related to some other specific detail of the particular system/configuration that was involved.

It'd be interesting to know if the issue reproduces with any other printers, or if it's limited to a certain driver/model.

You need to log in before you can comment on or make changes to this bug.