software-store-developers team mailing list archive

[Gary Lasker <gary.lasker@xxxxxxxxxxxxx>]

Hey Matt, I haven't actually had a chance to look at the hang situation
further yet today, but I sounds like you are on the right track. I tried
adding a main_iteration in the section that you mention (inside the
append_placeholders method) and indeed the UI is no longer blocked. The
only problem is that if I navigate to a details view and back to the
list, the list then becomes blank (and I see lots of "Interrupting ..."
messages being spewed).  But maybe I'm not doing things in the right
place, etc.  I don't think you've uploaded this change to your own
branch, correct?  At least, I'm not seeing it?  Anyway, if you do, just
ping me or send email; I'm happy to test!

Thanks!  And cheers,
Gary


On 02/25/2011 03:49 PM, Matthew McGowan wrote:
> Hey Gary!
> 
> I have not looked at your branch just yet, so you may have fixed this
> already, but i think i know what is causing the issue you are seeing.   
> 
> In LitsStore.set_from_matches i do the following:
> 
> # append empty rows
> for i in range(len(matches[initial_batch:])):
>      self.append()
> 
> On my system to append all the empty rows for the System cat takes 0.5
> sec.  Now in this loop I do main_iterations, this adds a bit of time to
> the whole build of the list but it probably solves the issue you are
> seeing. 
> 
> 
> Matt
> 
> On Fri, Feb 25, 2011 at 6:18 PM, Gary Lasker <gary.lasker@xxxxxxxxxxxxx
> <mailto:gary.lasker@xxxxxxxxxxxxx>> wrote:
> 
> 
>     Heya Matt and everybody,
> 
>     This branch feels like it's coming into shape. Things seem to be mostly
>     working for me now, but I am still seeing some strange pauses at times
>     and, in one case still, a complete freeze. This last happens when I
>     click on the "System" category. I see the list get rendered, but then Le
>     Centre just freezes up and clicking on things does nothing. Finally, the
>     app greys out.
> 
>     Note that this happens on my work laptop only. If I try this on my Dell
>     Mini, there is no freeze and the System category works fine. Hrm hrm.
> 
>     I also noticed a bug where clicking between the full "Provided by
>     Ubuntu" view and the installed "Provided by Ubuntu" would actually not
>     refresh the view. I have a potential fix for this in my branch:
> 
>      lp:~gary-lasker/software-center/appstore-experiment
> 
>     I'll dig more into the slowdowns and my "System" freeze issue tomorrow.
> 
>     Rock on developers!
>     Gary
> 
> 
>     On 02/24/2011 05:00 PM, Matthew McGowan wrote:
>     >
>     >
>     > On Fri, Feb 25, 2011 at 7:15 AM, Michael Vogt <mvo@xxxxxxxxxx
>     <mailto:mvo@xxxxxxxxxx>
>     > <mailto:mvo@xxxxxxxxxx <mailto:mvo@xxxxxxxxxx>>> wrote:
>     >
>     >     On Thu, Feb 24, 2011 at 08:57:14AM +1300, Matthew McGowan wrote:
>     >     > Hi folks,
>     >     Hi Matthew,
>     >
>     >     > I think i just pushed the last major bits of functionality
>     for my
>     >     appstore
>     >     > listmodel/treemodel branch.  I think it's working fine.  but
>     i am
>     >     a lazy
>     >     > tester.
>     >     >
>     >     > lp:~mmcg069/software-center/appstore-experiment
>     >
>     >     I worked on my version of this branch today:
>     >     lp:~mvo/software-center/appstore-experiment
>     >
>     >     I really like the new code layout, very nice aand smooth. Also the
>     >     delayed loading is pretty cool. I updated the tests and did some
>     >     cleanups. However I get a hang when I run "test_gui.py". Its 100%
>     >     reproducable for me (on three machines) and I wasn't able to
>     pinpoint
>     >     the issues. If I remove the test_gui.py
>     test_previous_purchase() all
>     >     tests pass. But the hang makes me nervous as it appears to be
>     >     something in the idle/timeout handler within gtk. My worry is
>     that its
>     >     something that is a deeper issue with the ListStore.
>     >
>     >     > The installedview particularly could use some brutal testing.  I
>     >     have now
>     >     > implemented filter and show|hide technical software.  Seems to
>     >     work nice
>     >     > enough, and I see no crashers or other heinous side-effects but
>     >     like i said
>     >     > i am a lazy tester.
>     >     >
>     >     > There are appview aspects which need polish, i.e. keyboard
>     >     expand-collapse
>     >     > of categories in the installedview.  There is prob more i have
>     >     overlooked,
>     >     > but they should be small UI things.  I also understand from mvo
>     >     some tests
>     >     > are failing.  I have not looked into this just yet, probably
>     tonight.
>     >
>     >     Those are updated now and work except for the one hang that I
>     outlined
>     >     above.
>     >
>     >     > If people get a chance, could you play with it see how it works
>     >     for you?
>     >     >  obviously feedback and/or code appreciated!  I imagine the
>     >     chances are slim
>     >     > right now but I'd love to see this in s-c tomorrow!  :P
>     >
>     >     I (and you too of course) worked hard today to make this
>     happend. I
>     >     still would love to merge it and get a feature-freeze
>     exception for it,
>     >     but I really want to get to the bottom of the hang before
>     doing that.
>     >
>     >     Cheers,
>     >      Michael
>     >
>     >
>     > mvo, thanks so much for your hard work yesterday, i appreciate the
>     help!
>     > hope it didn't get in the way of too much.
>     >
>     > I'll look into the hang and other things this weekend.  I suspect it's
>     > due to either overlapping appends and/or queries and things becoming
>     > inconsistent.
>     >
>     > I tried a decorator which interrupts and waits for an idle append to
>     > return, that seems quite effective and i think solves a subset of
>     errors
>     > i've seen.
>     >
>     > I am getting some xapian.DatabaseCorruption Errors in the branch which
>     > occur during either documt selection from the db
>     > (xapiandb.get_document(match.docid)) or on get_mset().  I dont really
>     > know what to think here.  Perhaps creating new AppEnquire's each
>     time we
>     > need to do a query, similar to how we use the existing AppStore?
>     >
>     > --
>     > From the mind of me!
> 
> 
> 
> 
> -- 
> From the mind of me!