subunit

invalid content types trigger a parser exception

Bug #1130280 reported by Clark Boylan
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
subunit
Triaged
Wishlist
Unassigned

Bug Description

OpenStack Nova is using testr and python subunit.run to run tests. The traceback at https://bugs.launchpad.net/nova/+bug/1130146/comments/1 has recently been observed (Invalid MIME type in the subunit stream processing). This appears to be the result of a change to the tests that are writing extra information to stdout (there is some thought that the tests themselves are broken).

Subunit should handle exceptions like this more gracefully. testr is seeing the tests as passing successfully. At the very least subunit should report an error in the subunit stream so that bugs like this in tests are caught early. It would be even better if subunit could recover and properly report results for the tests that did not inject extra bits into the subunit stream.

Full jenkins console logging and a subunit stream for a failed test run can be found at http://logs.openstack.org/22330/1/check/gate-nova-python27/16956/.

Revision history for this message
Robert Collins (lifeless) wrote :

Arguably the parser excepting is ok, except that due to a bug in testtools ThreadSafeForwardingResult/ConcurrentTestSuite, the exception is lost.

Changed in subunit:
status: New → Triaged
importance: Undecided → Critical
summary: - Python Subunit should more gracefully handle unexpected subunit stream
- contents.
+ invalid content types trigger a parser exception
Revision history for this message
Robert Collins (lifeless) wrote :

https://bugs.launchpad.net/subunit/+bug/1130280 tracks the subunit parsing side of this; I don't believe there are any code changes for testrepository itself, so I'm going to remove that task. I will file a separate bug for testtools for the failure to notify of the processing error.

Revision history for this message
Robert Collins (lifeless) wrote :

BAH, wrong bug.

Revision history for this message
Robert Collins (lifeless) wrote :

bug-1130429

Revision history for this message
Robert Collins (lifeless) wrote :

Dropped to critical - reasoning: the testtools concurrency driver ensures failures will be documented and not silently hidden, so while we should do this it isn't a major issue.

Changed in subunit:
importance: Critical → Wishlist
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.