migration-assistant cleanup handler does very weird stuff involving recursively calling run_command
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
ubiquity (Ubuntu) |
Fix Released
|
Medium
|
Evan |
Bug Description
Binary package hint: ubiquity
During an installation test here, I get a warning dialog saying "MigrationAssistant failed with exit code 32". However, the log indicates that ma-ask exited 0. Baffled, I added some further debugging to investigate, and discovered that something in the middle of the migration-assistant component's cleanup handler is forking! (I happened to add a debug statement after the cleanup handler, and I get one line of output from the pre-cleanup debug and two lines of output from the post-cleanup debug.) In the process no doubt it's trashing the local variable containing the return code, but more importantly I have absolutely no idea where the control flow is going to go afterwards.
This is probably due to the stuff in MigrationAssist
Changed in ubiquity: | |
importance: | Undecided → High |
status: | Unconfirmed → Confirmed |
Changed in ubiquity: | |
assignee: | nobody → evand |
Judging from stack traces, this is because ok_handler calls run_command again. Ugh.