1. Impact: Significant impact on heavily loaded or under-powered systems; can take several seconds (up to a minute) to launch a screen session; these fixes can improve this by an order of magnitude
2. How fixed: The fix drops the ionice and cpu nice calls which wrap all status scripts; this was added very late in the Jaunty release and is not desirable; cause screen to take a very long time to launch; also, we eliminate a couple of very expensive calls to lsb_release; finally, we drastically improve the most expensive status item (updates-available) by backgrounding the expensive run and writing that output to the cache file which we already read
3. Patch: in the debdiff
4. Test Case: Run 'time screen true'. On low power systems (like arm), or on heavily loaded systems, this can take 5 seconds - 60 seconds. Simply unacceptable. With this upload, all but the slowest systems should be able to launch screen within a couple of seconds (most taking less than a second).
5. Regression potential: Low. I have tested this thoroughly. The ionice and cpu nice calls were a last minute addition and a poor choice; they don't actually do what we want. It's better to just let the system's normal scheduling mechanisms handle this.
SRU Justification:
1. Impact: Significant impact on heavily loaded or under-powered systems; can take several seconds (up to a minute) to launch a screen session; these fixes can improve this by an order of magnitude
2. How fixed: The fix drops the ionice and cpu nice calls which wrap all status scripts; this was added very late in the Jaunty release and is not desirable; cause screen to take a very long time to launch; also, we eliminate a couple of very expensive calls to lsb_release; finally, we drastically improve the most expensive status item (updates-available) by backgrounding the expensive run and writing that output to the cache file which we already read
3. Patch: in the debdiff
4. Test Case: Run 'time screen true'. On low power systems (like arm), or on heavily loaded systems, this can take 5 seconds - 60 seconds. Simply unacceptable. With this upload, all but the slowest systems should be able to launch screen within a couple of seconds (most taking less than a second).
5. Regression potential: Low. I have tested this thoroughly. The ionice and cpu nice calls were a last minute addition and a poor choice; they don't actually do what we want. It's better to just let the system's normal scheduling mechanisms handle this.
:-Dustin