Mir

[regression] Fatal DRM errors from a compositing thread have no usable stack trace

Bug #1316867 reported by Daniel van Vugt
22
This bug affects 3 people
Affects Status Importance Assigned to Milestone
Mir
Won't Fix
High
Unassigned
0.4
Won't Fix
High
Unassigned
0.5
Won't Fix
High
Unassigned

Bug Description

Fatal DRM errors from a compositing thread have no usable stack trace. This is a regression of bug 1237332 caused by this landing:
    https://code.launchpad.net/~afrantzis/mir/fix-1189770/+merge/217652

An example of the problem is seen in bug 1294510 where the available stack traces are completely useless.

Related branches

summary: - [regression] Fatal exceptions raised in a compositing thread have no
- usable stack trace
+ [regression] Fatal DRM errors from a compositing thread have no usable
+ stack trace
description: updated
tags: added: regression
Changed in mir:
status: Triaged → In Progress
description: updated
Changed in mir:
milestone: 0.2.0 → 0.3.0
Changed in mir:
milestone: 0.3.0 → 0.4.0
Changed in mir:
milestone: 0.4.0 → 0.5.0
tags: added: rtm14
Changed in mir:
milestone: 0.5.0 → 0.6.0
tags: removed: rtm14
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

The team has voted that exceptions are preferred over the ability to debug crash reports, so this bug is not getting fixed any time soon. Maybe we can find middle ground in future.

Changed in mir:
status: In Progress → Triaged
milestone: 0.6.0 → none
assignee: Daniel van Vugt (vanvugt) → nobody
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Argh, this is annoying. No stack traces from fatal errors any more.

Revision history for this message
Alan Griffiths (alan-griffiths) wrote :

Have you tried running with --on-fatal-error-abort?

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Don't care for now...

Although isn't that the default option? If not then it should be. Otherwise crash reports from real users will be unreadable.

Revision history for this message
Alan Griffiths (alan-griffiths) wrote :

As discussed at length over a year ago Mir is a library and, as such, should not impose that decision all applications using it. The default is to report the failure using an exception (as per comment #1).

For downstream servers we recommended that to get good crash information they consider selecting the "abort" behaviour.

Changed in mir:
status: Triaged → Won't Fix
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers