Red Hat Bugzilla – Bug 440659
search should cancel current search operation
Last modified: 2008-04-07 22:00:49 EDT
Description of problem:
The packagekit pk-application install UI should cancel any prior search
operation when a new one is started. If you currently enter a search term and
hit enter, then immediately change the search term and hit enter, and repeat
this, then multiple searches will be queued which will all complete one at a
time (or it may just lock up...).
It is not necessary to attempt to finish a search operation if a new one was
started, because it will have nowhere to be displayed (this is not a tabbed or
multiple view application).
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. start several searches in a row before they are able to complete
Created attachment 300425 [details]
showing multiple running searches
Does this still happen if you use the packages here:
No, it looks like I cannot cause this with:
At most I see two operations at once shown in the context menu. This was most
noticeable on my x86_64 virtual machine which runs everything slower and may
still react a bit less quickly.
My screenshot above was only showing 2 entries, but I had seen up to 5 of them
left over 'waiting for..', so now seeing max 2 is different at least.
How are you managing to get two queued searches? Do you have two instances of
Nope, open one instance, then start a search, change the search term, hit enter, change the search term,
hit enter, etc. If the network happens to stall just a little bit (i.e. the search takes more than a few
seconds, most do) then the searches start to stack up.
I found this when I typed the search term wrong several times accidentally, fat fingering my searches a
Note that it was actually slowing down the final result, not just having multiple entries show up in the
context menu on the panel applet. On my x86_64 virtual machine I was seeing the 5 or 6 entries, which
then each disappear one at a time (finished searching) until the final search would run and show me the
results. As I said with your test repo (on my i686 machine) I don't see this happen, instead they appear to
be trying to cancel and run only the last entered search. I don't know if its due to the network/machine
performance difference or if there is a code difference that should show that result.
Hmm. Could you please attach the *complete* output of:
when you manage to get the multiple queue. Thanks.
Ok I can't reproduce this on PK-0.1.11-1, either x86_64 or i386. There are 2
entries shown in the applet context menu only long enough to notice if I try to
change the search term and immediately click the applet to see the menu.
Repeated searches appear to be canceling prior searches quickly now. I'm
attaching a long output of things working correctly.. except for the broken
access to database several times at the beginning, then after refresh
application list it starts working.
I'm going to close this as it seems to be working for now, but attachment coming
Created attachment 301590 [details]