This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
Performance tests consistently show a regression in times of displaying Update Center Wizard dialog box. The absolute times together with the percentages of comparison to NB351: Internationalize dialog open (1st invocation) linux: 1 578 ms (+81%) solaris: 2306 ms (+76%) xp: 1 417 ms (+77%)
Jirka, sorry :(
Ups, sorry my mistake Correct numbers are : Update Center Wizard dialog open (1st invocation) linux: 1 578 ms (+81%) solaris: 2306 ms (+76%) xp: 1 417 ms (+77%) by the way second invocation is 20-30% faster than NB3.5.1 :)
Am I blind ? What the hell is difference between these and those numbers ?
Lukas, wake up :) caption is different - not numbers :)
yes, I've noticed the different caption with the same numbers ;) but I was suprised why you are talking about numbers... ale kurna, who cares - everythings ok, now, let's stop this stupid disscusion and let Jirka fix it :)
The time of first invocation AU is much better, no regression now. Improvement was caused by change autoupdate/../Wizard, rev.1.64 (Log: improved AU Wizard's performance, postponed creating Swing components as late as possible). I keep this issue open to track another suspected places participated in the regression, feel free to close it anytime.
Confirming that we don't measure the regression any more. Feel free to close this bug.
Execution NbPresenter.show() doesn't make a regression because is varied from each run to run, there's no need to fixed. The improvement which is already done is good enough, I closing as FIXED.
Jirka, I am sorry, but there is still regression , numbers from 200402221900 : Linux 1 370 ms (+59.5%) Sol 9 807 ms (-33.4%) WinXP 1 261 ms (+60.1%) reopen Don't worry, we are waiting for today's numbers, maybe it's once time regression ;)
I don't agree I don't do any changes which could affect processing open AutoUpdate. I run the MeasureDialogs tests against today's sources and I can see no regression. I'm going to close it unless a next run your tests confirm a regression.
There are the tests results from the three last build, which weren't any changes in code: build: 2/24 2/22 2/19 winXP: 1 589 ms (+25.6%) 1 261 ms (-0.3%) 1 239 ms (-2.0%) solaris: 2 103 ms (+39.0%) 807 ms (-46.6%) 808 ms (-46.6%) linux: 737 ms (-41.9%) 1 370 ms (+8.0%) 621 ms (-51.0%) Because the results are very random, I close as FIXED. Reopen only if there is a evidence which integration caused the new regression.
To assure my report that there is no regression now I did next measurement Open AU wizard via watching EventQueue. I run nb351 and nb36 8 times, trash the best and the worst results and count average. Here are the averages: linux/Gnome WinXP 351 1337 683 36 985 636 Truth is, the times of each run differs from 500ms to 900ms on WinXP and from 600ms to 1400ms on linux. The end result, in my opinion, is, there is no regression after my fix from last week.
Ok, verifying