- User Since
- Jul 8 2015, 8:34 AM (106 w, 4 d)
Sat, Jul 22
Fri, Jul 21
Fri, Jul 7
Hmm... i was thinking this could cause issues, i don't see anywhere assumptions are made about page size on the consumption side (damn we're lucky this whole split thing's so new ;) ). Nothing in the API suggests you might get more than the requested number when you ask for entries, but conversely nothing says you will potentially get less, and that happens on a regular basis, so... sure, why not :)
Wed, Jul 5
Tue, Jul 4
Different approach, but yeah, telling people what'll actually happen works :)
derp, not accepted, my bad...
That is indeed a very good point. I think we might possibly have an issue here, though, in that since these two searches are no longer stand-alone, if the filter is not explicitly reset before attempting to perform a search we will only search installed and updateable items. I don't think it is necessarily an enormous or insurmountable issue here, as far as i can tell the only code which currently uses these two functions is in KNS' own DownloadManagers (at least, lxr suggests as much), and if we ensure those two are fixed to reset the filter before launching a search, i think we're probably ok.
Tue, Jun 27
Add/fix a whole bunch of more documentation, identified as missing (or incorrect) by Aniketh
Jun 23 2017
Jun 22 2017
Looks good for the initial case of getting rid of the obviously broken looking aspect ratio at the very least, though i think exposing the property might still be good... But, at the very least, it matches the HIG so that's cool :) (i don't personally like the whole scale-and-crop thing, but i'm also not the designer and i know entirely sensible reasons exist for this choice)
Jun 15 2017
Incidentally, because KNS itself will already postpone non-cached searches by one second, the new proposed solution above (to only to this for KNS sources) is already implemented by way of using D6191
Jun 14 2017
Related bug: https://bugs.kde.org/show_bug.cgi?id=380138
Upon reviewing prior to merging, i realised that a file was missing from the most recent diff (StandardBackendUpdater did, as suggested, not listen to the resourceRemoved signal, and now does).
Jun 13 2017
Fix glaringly obvious over-complication spotted by David :)
Adapted code as suggested. Note that this now depends on D6190 getting merged (as that contains the code that's iffed out).
Jun 12 2017
Don't const & an int, that's just silly.
As far as i can gather, it was simply never added because, well, it was never used in a lot of places... This really is more a case of equalising some features between requestData and other parts of the engine (so they can all be paginated by the size they really want to be).
Jun 8 2017
Jun 6 2017
On a similar note to handling comments, how does it now handle unknown/garbage tags? While it won't affect the cache code, it would potentially affect other things (ocs is not guaranteed to be perfectly formed, and it's one of the ways the framework's retained backwards compatibility). From what i can tell, this would assert when an unknown tag is encountered, right?
Personally less fond of auto than you are... But, that is just me, and this is a framework, and it's fine :) A massive reduction in allocations is a very good thing, go for it :)
Jun 5 2017
Sorting correctness (and more thorough testing) is good, yes. LGTM! :)
May 18 2017
May 17 2017
May 15 2017
May 11 2017
May 8 2017
Have a slightly more useful fallback icon than just nothing (that is, if there is no first thumbnail, use the one it used previously)
May 7 2017
Work some numbers a bit
Static var naming change, for consistency and whatnot
static var naming fix, for consistency with elsewhere
A bit of naming cleanup, and a very silly miscalculation because things count in bytes rather than some arbitrary random multiple of them
Simplify the xmlloader cache logic a touch
Some style fixes, and set a reasonable maximum size for the cache
May 5 2017
Unpointerify the internals, as agreed