1663536 - Large increase in printing errors reported via telemetry

Status: RESOLVED FIXED

(Core :: Printing: Setup, defect, P1)

Description

[Jonathan Watt [:jwatt]]

The telemetry graphs are in flux as Martin has been reworking the queries to try to get us more actionable data, but it appears there is a large increase in NS_ERROR_FAILURE printing errors being reported on Windows in Nightly around 20200826214121. From what I can tell it looks like the regression occurred between 2020-08-25-15-44-12-mozilla-central and 2020-08-26-21-41-21-mozilla-central, which gives this regression range:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=a8f04609cdd89e6dd4f6e8c3f24daca3b6dc33be&tochange=61ed3192760a3aaef282c089c064fc8dd3890125

It's a little hard to tell what's going on right now though. The total error graphs have been edging up since around 7th August. The date based graphs displayed up until earlier today were showing around a 5x increase. The current build_id based graphs seem to be a bit broken right now and aren't so clear.

The large increase in errors is present on Beta too.

We should probably wait until Martin has managed to wrangle useful data out of SQL tomorrow, but I'm filing this to get in or the radar to make sure it isn't overlooked.

Comment 1

[Sean Voisen (:svoisen)]

This appears to have died down since 81.0b7. I guess keep this open a few more days ...

Comment 2

[Jonathan Watt [:jwatt]]

Unfortunately, the reduction in errors reported on Nightly is due to bug 1664129.

I don't think we can infer anything from the reduction in error rate on Beta due to the pref auto-disabling as of b7/2020090616 (other than that the error rate is not increased when the old UI is being used, which is good!).

Comment 3

[Jonathan Watt [:jwatt]]

Scrap that, the reporting's fine. It was just too early in the morning and I was forgetting the details of the indirect way we do the reporting.

Comment 4

[Jonathan Watt [:jwatt]]

GFX_PRINTER_DOC_IS_BUSY appears to have essentially stopped happening in Nightly between 2020-07-06-09-51-44 and 2020-07-06-21-50-07. From the commits that landed, bug 1636803 seem like the only one that might have fixed this error.

NS_ERROR_NOT_AVAILABLE appears to have essentially stopped happening in Nightly between 2020-08-25-15-44-12 and 2020-08-26-03-24-37. From the commits that landed, bug 1658819 (commit message "Wait for preview to finish before updating page size") and bug 1636728 seem like the most likely fixes.

Regarding NS_ERROR_FAILURE, the spike in which this bug was really filed for, the error rate seems to have really jumped between 2020-09-03-15-18-16 and 2020-09-04-09-43-41. From the commits that landed, bug 1662975 and bug 1663053 seem like the only relevant bugs. The error rate then dropped back down to "normal" in Nightly between 2020-09-06-09-41-18 and 2020-09-06-21-50-57. From the commits that landed, bug 1663283, bug 1663140 and bug 1662259 seem like the suspect bugs. Without any deeper analysis, since bug 1662975 was followed up by regression fix bug 1663140, I suspect they're the likely cause and fix for the spike in these errors.