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.
Automatic performance tests consistently show a regression in the times for opening editor for properties, TXT and XML files on Windows XP. (Linux and Solaris do *not* have a regression.) The absolute numbers together with the percentage of comparison to NB351: Open Properties file (20kB) (1): 719 ms (+10.1%) Open Properties file (20kB) (2): 438 ms (+26.9%) Open Properties file (20kB) if Editor opened (1): 354 ms (+10.0%) Open Properties file (20kB) if Editor opened (2): 312 ms (+16.6%) Open Txt file (20kB) (1): 651 ms (+57.9%) Open Txt file (20kB) (2): 545 ms (+60.7%) Open Txt file (20kB) if Editor opened (1): 401 ms (+20.0%) Open Txt file (20kB) if Editor opened (2): 422 ms (+30.0%) Open Xml file (20kB) (1): 859 ms (+35.5%) Open Xml file (20kB) (2): 580 ms (+57.8%) Open Xml file (20kB) if Editor opened (1): 417 ms (+11.1%) Open Xml file (20kB) if Editor opened (2): 460 ms (+28.9%)
I would like to ask the perf team to measure opening times after the integration of the new windowing system prior to integration of the code-folding enabled editor i.e. trunk builds from e.g. 2003/12/15. If those builds would already show the regression then it would be fair to reassign this issue to core/window-system so that they run the Performance analyzer and do the complete analysis of the file opening. I do not see any reason why the editor team should be responsible for analyzing such complex thing as file opening where many components of the IDE interact. I have no idea whether we will be able to improve the numbers significantly especially if they are affected by the are line view objects created. Setting target milestone to 3.6 but there is potential for waving of this issue.
Mila, I ran these tests on my machine (Windows 2000) with builds 20031215 and 20040217. Some of the numbers are strange (-60%, -57%, 0%) but the rest of the numbers show a certain level of regression between these two builds. I ran the test suite twice on both builds and the results were almost the same each time. Please, note that the absolute values are not be comparable with the numbers above, just the percentages are relevant... [TEST NAME]: [value w/ 20040217] ([comparison to 20031215]) Open Properties file (20kB) (1): 828ms (+8%) Open Properties file (20kB) (2): 302ms (+45%) Open Properties file (20kB) if Editor opened (1): 328ms (-60%) Open Properties file (20kB) if Editor opened (2): 281ms (+23%) Open Txt file (20kB) (1): 750ms (-57%) Open Txt file (20kB) (2): 473ms (+14%) Open Txt file (20kB) if Editor opened (1): 438ms (+17%) Open Txt file (20kB) if Editor opened (2): 406ms (+11%) Open Xml file (20kB) (1): 1031ms (0%) Open Xml file (20kB) (2): 468ms (+40%) Open Xml file (20kB) if Editor opened (1): 438ms (+40%) Open Xml file (20kB) if Editor opened (2): 422ms (+47%)
Requesting Bug Waiver for 3.6 Justification --------------------- The action is still in resposivness limit (<1s) for 20KB file, measured on Pentium 700Mhz.
Waiver approved.
According to Marial Mirilovic, there is no regression in the latest measurements.
Tondo, could you please verify this issue? Thanks.