RESOLVED FIXED 234107
'run-benchmark' should launch browsers in a relative clean state.
https://bugs.webkit.org/show_bug.cgi?id=234107
Summary 'run-benchmark' should launch browsers in a relative clean state.
dewei_zhu
Reported 2021-12-09 15:31:26 PST
'run-benchmark' should launch browsers in a relative clean state.
Attachments
Patch (14.41 KB, patch)
2021-12-09 15:46 PST, dewei_zhu
slewis: review+
dewei_zhu
Comment 1 2021-12-09 15:46:37 PST
dewei_zhu
Comment 2 2021-12-09 16:51:50 PST
Stephanie Lewis
Comment 3 2021-12-10 13:58:04 PST
Comment on attachment 446624 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=446624&action=review > Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_chrome_driver.py:34 > + chrome_options.add_argument("--disable-extensions") did you want disable-extensions in two places? > Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_chrome_driver.py:41 > + return '--window-size={width},{height}'.format(width=int(screen_size.width), height=int(screen_size.height)) does setting window size set content size or window size? > Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_firefox_driver.py:33 > + args_with_url = ['--args', '-width', str(int(screen_size.width)), '-height', str(int(screen_size.height))] Do we really run fullscreen? Same question does this set content size or window size because Chrome, Firefox, and Safari often have different sized chromes and we want the content size to match
dewei_zhu
Comment 4 2021-12-12 17:52:50 PST
Comment on attachment 446624 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=446624&action=review >> Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_chrome_driver.py:34 >> + chrome_options.add_argument("--disable-extensions") > > did you want disable-extensions in two places? Those are different, one for webdriver, one for launch with 'open' command. >> Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_chrome_driver.py:41 >> + return '--window-size={width},{height}'.format(width=int(screen_size.width), height=int(screen_size.height)) > > does setting window size set content size or window size? It will fill the screen, that's what we want. >> Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_firefox_driver.py:33 >> + args_with_url = ['--args', '-width', str(int(screen_size.width)), '-height', str(int(screen_size.height))] > > Do we really run fullscreen? > Same question does this set content size or window size because Chrome, Firefox, and Safari often have different sized chromes and we want the content size to match This will not be full screen, it will fill screen without overlapping with docker.
dewei_zhu
Comment 5 2021-12-16 16:06:09 PST
Comment on attachment 446624 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=446624&action=review >>> Tools/Scripts/webkitpy/benchmark_runner/browser_driver/osx_chrome_driver.py:41 >>> + return '--window-size={width},{height}'.format(width=int(screen_size.width), height=int(screen_size.height)) >> >> does setting window size set content size or window size? > > It will fill the screen, that's what we want. Discussed in person. It looks like MotionMark is immune from slightly window height change and speedometer2 doesn't show significant change (100 iteration) with slightly different window size.
dewei_zhu
Comment 6 2021-12-20 17:20:24 PST
Landed in r287289.
Note You need to log in before you can comment on or make changes to this bug.