Shawn Walker wrote: > On 05/12/11 02:13 PM, Danek Duvall wrote: > >Shawn Walker wrote: > > > >> http://cr.opensolaris.org/~swalker/pkg-composite-2/ > > > >catalog.py: > > > > - line 3411: this disallows ver == 0, if I'm reading it correctly. > > Since it is always supposed to be a string (per docstring), no.
Okay; missed that. > But even then, 'ver' will never be just 0 for our use (it'll always be > something like 1.0,5.11, etc.): I've used @0 as a test quite frequently. > >publisher.py: > > > > - line 1238,1239: First "list", then "set", plus the name of the method? > > It doesn't mean "set" literally (as in the data type). This is the > logical sort of set; perhaps 'combo' instead of 'set'? I think a bit more consistency in the docstring is all that's needed here. > > - line 1313: urlencoding instead of hashing the uri would make for easier > > debugging. Hm. Maybe move the comment on line 1319 up here? And > > perhaps only hash if the name is too long? I doubt it'll happen too > > often, and there won't be any conflict. > > I'd rather not go there; it is preferable to me to be consistent and > avoid the path length issues entirely. I'd be willing to write out > the URI into a file in the directory for debugging purposes though if > it's really that important. All right; I wouldn't worry about it, then. > > - line 1487: I don't see how this could ever be True. It would have to > > be either None or an empty dict, and in either case, the tests on lines > > 1473, 1479, or 1484 would have blown up. > > Yes, it could be an empty dict in the event that all sources provide > a package (line 1473) and they are all V1 sources (line 1479). At > that point it *becomes* an empty dict -- it didn't start that way. > This simply reduces the amount of data that gets serialised. Gotcha. Thanks, Danek _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
