The rpmtsCallback() needs to be eliminated
The rpmtsCallback() used for progress/status communication
was originally intended to feed CPU cycles to a GUI installer.
As additional applications -- particularly bindings -- have
been written, there are complex side-effects driven by
events sent from the synchronous (i.e. rpm lib ceases to
run) events delivered through the callback.
Meanwhile the main functions of rpmtsCallback() are quite simple:
pushing progress bars
delivering state information (like what pkg is installed)
that could be done by sharing a state structure instead.
The more complex operations of error handling SHOULD be done
through an exception framework, not as events delivered through
a synchronous callback mechanism.
Blueprint information
- Status:
- Not started
- Approver:
- Jeff Johnson
- Priority:
- High
- Drafter:
- Jeff Johnson
- Direction:
- Approved
- Assignee:
- Jeff Johnson
- Definition:
- Drafting
- Series goal:
- Accepted for 5.4
- Implementation:
- Not started
- Milestone target:
- None
- Started by
- Completed by
Related branches
Related bugs
Bug #635854: Need better handling of no space left rpm errors | Triaged |
Bug #637173: Forgotten RPMCALLBACK_INST_STOP? | New |
Bug #912429: RPMCALLBACK_INST_STOP added by rpm.org | New |
Sprints
Whiteboard
Work Items
Dependency tree
* Blueprints in grey have been implemented.