General comments: Could you please update the bugs with the analysis you've done here and your answers to Padraig's questions? I think that would be helpful.
I'm not quite sure where the speedup is coming from in lines 1470-1505. Is it because pkgs_known becomes a much smaller list? I'd be interested in seeing the output of /opt/DTT/Python/py_calltime or py_cputime to see where the speed up is happening, and where the continuing bottle necks are. Lines 1470-1505: I'd rather this code wasn't copied, we've been down that road before. If the gui needs it as well, then I'd suggest promoting it and placing it in either misc or api.py. Which one, I don't really feel strongly about. nits: 956: please indent this correctly Brock Michal Pryc wrote: > Padraig O'Briain wrote: > >> I had a few questions and I sat with Michal and got answers for them. >> >> Q. What changes were made for 4341? >> A. The use of state is get_image_from_directory ensure that GUI results >> are the same as CLI results >> >> Q. How is the same package being in more than one authority handled? >> A. The use of pkg_stem ensures that we have separate rows for the two >> packages. >> >> One minor nit: >> The assignment of default_auth in update_package_list is not necesary as >> it is unused. >> > Padraig, > Thanks, the webrev with fix for minor nit: > > http://cr.opensolaris.org/~migi/05_12_2008_bug5645_5289_4341_5150_v1 > > Michal > > >> Padraig >> >> >> On 12/05/08 11:31, Michal Pryc wrote: >> >>> Hello, >>> The webrev is at: >>> http://cr.opensolaris.org/~migi/05_12_2008_bug5645_5289_4341_5150/ >>> >>> Fixes: >>> 1. >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=5645 >>> Package Manager should have busy cursor for certain actions >>> >>> 2. >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=5289 >>> Package Manager do not respect the same packages from non default >>> authorities >>> >>> 3. >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=4341 >>> IPS GUI and CLI list different status for some packages >>> >>> 4. >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=5150 >>> PM loading of package entries takes too long >>> >>> The 4. is partialy fixed. The time is reduced on my machine from: >>> >>> After __init__ : 0.31704211235 >>> Call to load_catalogs: 10.3101949692 >>> Enter get_list func: 10.3106389046 >>> After call to sort: 17.8955469131 >>> Return get_list func: 39.3143429756 >>> >>> As we can see the total time before was 39 seconds for returing from >>> the get_list func which is basically the last func for startup. >>> >>> The new times (proposed in this webrev): >>> >>> After __init__: 0.258629083633 >>> Call to load_catalogs: 10.1712219715 >>> Enter get_list func: 10.171653986 >>> After call to sort: 15.8203570843 >>> Return get_list func: 17.8212211132 >>> >>> As we can see we are up and running in 17.8 sec which is ~45% of the >>> time spend before improvements. The bottle neck is the call to >>> img.load_catalogs() which is more then 60% of the time for the startup. >>> >>> best >>> Michal >>> >> _______________________________________________ >> pkg-discuss mailing list >> [email protected] >> http://mail.opensolaris.org/mailman/listinfo/pkg-discuss >> > > _______________________________________________ > pkg-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/pkg-discuss > _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
