Closed Bug 1793978 Opened 2 years ago Closed 1 year ago

Crash in [@ mozilla::a11y::Accessible::IsHTMLBr]

Categories

(Core :: Disability Access APIs, defect)

Desktop
Linux
defect

Tracking

()

RESOLVED FIXED
109 Branch
Tracking Status
firefox-esr102 --- unaffected
firefox107 --- wontfix
firefox108 --- wontfix
firefox109 --- fixed

People

(Reporter: snegritas, Assigned: Jamie)

References

(Regressed 1 open bug)

Details

(Keywords: crash, Whiteboard: [ctw-m4])

Crash Data

Attachments

(1 file)

Crash report: https://crash-stats.mozilla.org/report/index/8b69f935-37e2-4c82-aa46-85a760221006

Reason: SIGSEGV / SEGV_MAPERR

Top 10 frames of crashing thread:

0 libxul.so mozilla::a11y::Accessible::IsHTMLBr const accessible/basetypes/Accessible.h:493
0 libxul.so mozilla::a11y::TextLeafPoint::IsEmptyLastLine const accessible/base/TextLeafRange.cpp:506
1 libxul.so mozilla::a11y::TextLeafPoint::FindBoundary const accessible/base/TextLeafRange.cpp:858
2 libxul.so mozilla::a11y::HyperTextAccessibleBase::TextAtOffset accessible/basetypes/HyperTextAccessibleBase.cpp:495
3 libxul.so mozilla::a11y::RemoteAccessible::TextAtOffset accessible/ipc/other/RemoteAccessible.cpp:226
4 libxul.so getTextAtOffsetCB accessible/atk/nsMaiInterfaceText.cpp:208
5 libatk-1.0.so.0 atk_text_get_text_at_offset atk/atktext.c:499
6 libatk-bridge-2.0.so.0 impl_GetTextAtOffset atk-adaptor/adaptors/text-adaptor.c:177
7 libatk-bridge-2.0.so.0 handle_message droute/droute.c:601
8 libdbus-1.so.3 dbus_connection_dispatch /build/dbus-IAhSsk/dbus-1.12.20/dbus/dbus-connection.c:4576

I have managed to reproduce the issue with firefox 107.0a1(2022-10-06) on Ubuntu 22.04 while browsing with keyboard navigation on Google Drive. This was a random crash and it happened only once. I don't have any steps on how to reproduce it I will keep an eye if it happens again and try for some steps.

Jamie can this crash be something related to the "Cache the World" feature?

Flags: needinfo?(jteh)

This is related to CTW, yes. Thank you.

Blocks: a11y-ctw
Component: Disability Access → Disability Access APIs
Flags: needinfo?(jteh)
Product: Firefox → Core

Morgan, if you have any ideas on this one, I'd be keen to hear them... but it looks like another weird "how on earth could this be null" bug.

Flags: needinfo?(mreschenberg)

Blah. Looked for a bit, but no ideas here. So, so odd.

Flags: needinfo?(mreschenberg)

I just have seen occurence of this.
I was responding to a ticket at our internal instance of Atlassian Jira cloud.
I have mentioned another user, selected the whole line and received the crash.
I can't find a real repro steps at the moment.

Oh. I think the client might be passing an invalid start offset here. It looks like the ATK code nor HyperTextAccessibleBase verifies this. That would result in an invalid TextLeafPoint for start.

Whiteboard: [ctw-m4]

Playing with this more I think it must have something to do with embedded objects such as links or user mentions on the Jira cloud.
I have got two more crashes this morning.
I have put some comments what I think was happening at the time.
However when trying to reproduce this on github or with a CK Editor demo I can't.
Thus I am still unable to find a proper steps to reproduce.

An invalid aOffset results in an invalid TextLeafPoint, so just return early if that happens.

Assignee: nobody → jteh
Status: NEW → ASSIGNED
Pushed by jteh@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/aec61c34aca0
Fail gracefully (instead of crashing) when an invalid offset is given to HyperTextAccessibleBase::TextAt/Before/AfterOffset. r=morgan
Regressions: 1804184
Status: ASSIGNED → RESOLVED
Closed: 1 year ago
Resolution: --- → FIXED
Target Milestone: --- → 109 Branch

Well, I guess this does actually affect Fenix outside of Nightly, but the volume is low enough that it can ride the trains.

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

Attachment

General

Created:
Updated:
Size: