1587856 - Incorrect find in page results from GeckoSession.SessionFinder when invoked in quick succession

Status: NEW

(GeckoView :: General, defect, P2)

Description

[Christian Sadilek [:csadilek]]

While debugging intermittent find-in-page test failures in Fenix, I came across the following problems with GeckoSession.SessionFinder:

Incorrect results after clear():

• Open blog.mozilla.org
• Type "a": Calls find("a", 0) -> result e.g. 85/318
• Type "n": Calls find("an", 0) -> result e.g. 21/56
• Type quickly backspace, backspace (calls clear()), then "a" (calls find("a", 0)) -> expected similar result to first search of "a", but getting 0/0 with found = true)

Incorrect results on find next/previous:

• Open blog.mozilla.org
• Type "a": Calls find("a", 0) -> result e.g. 85/318
• Click next result a few times quickly: Calls find(null, 0) -> getting results with 0/0 instead of accurate current/total.

Comment 1

[Chris Peterson [:cpeterson]]

snorp says this looks like a race in the text finder code. P2 because this bug is reproducible manually, not just in the test.

Comment 2

[eclaudiu64]

Attachment: mobizen_20230927_163018.mp4

Search page, incorrect result appears, on android. I put the letter "c" and being very many, the result in the border appears as it is in the attached video.

Comment 3

[eclaudiu64]

(In reply to eclaudiu64 from comment #2)

Created attachment 9355324 [details]
mobizen_20230927_163018.mp4

Search page, incorrect result appears, on android. I put the letter "c" and being very many, the result in the border appears as it is in the attached video.

There are over 5000 results on the page, but the complete result appears on -1. This is the page where I noticed: https://bugzilla.mozilla.org/buglist.cgi?bug_status=__open__&component=Privacy%3A%20Anti-Tracking&list_id=16735601&product=Core&query_format=advanced&order=bug_id&limit=0, but it happens on any site where there are thousands of results.
Is there an update to this error?

Comment 4

[Chris Peterson [:cpeterson]]

"-1" results is a bug, but I don't have an update on the status of the fix at this time.