Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
--- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42281 --- Ship it! Looks ok. It will need testing with the startkde/startactive. - Ivan Čukić On Oct. 24, 2013, 2:04 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:04 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
--- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42283 --- This review has been submitted with commit 90e2c0b65cce3f899d4fd5514493beaa60151f1e by Martin Klapetek to branch master. - Commit Hook On Oct. 24, 2013, 2:04 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:04 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
--- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Status -- This change has been marked as submitted. Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
--- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. - Fredrik Höglund On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 6:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. - Martin --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 4:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 4:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. - Ivan --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. Ivan Čukić wrote: We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. Isn't the splash screen supposed to cover (also) those loading costs? If the splashscreen took like 5 seconds to load but then could finish in 500ms, we would simply not require a splash screen in the first place. - Thomas --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. Ivan Čukić wrote: We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. Thomas Lübking wrote: Isn't the splash screen supposed to cover (also) those loading costs? If the splashscreen took like 5 seconds to load but then could finish in 500ms, we would simply not require a splash screen in the first place. I've never done any (useful) startup benchmarks, but I'd be happy to try. Can you suggest some ways to measure cold startup time? Then I'll test both and post my findings. - Martin --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. Ivan Čukić wrote: We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. Thomas Lübking wrote: Isn't the splash screen supposed to cover (also) those loading costs? If the splashscreen took like 5 seconds to load but then could finish in 500ms, we would simply not require a splash screen in the first place. Martin Klapetek wrote: I've never done any (useful) startup benchmarks, but I'd be happy to try. Can you suggest some ways to measure cold startup time? Then I'll test both and post my findings. If loading Qt took 5 seconds, loading the rest would take 10 minutes. As I said doesn't seem to be noticeably slower. - Ivan --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. Ivan Čukić wrote: We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. Thomas Lübking wrote: Isn't the splash screen supposed to cover (also) those loading costs? If the splashscreen took like 5 seconds to load but then could finish in 500ms, we would simply not require a splash screen in the first place. Martin Klapetek wrote: I've never done any (useful) startup benchmarks, but I'd be happy to try. Can you suggest some ways to measure cold startup time? Then I'll test both and post my findings. Ivan Čukić wrote: If loading Qt took 5 seconds, loading the rest would take 10 minutes. As I said doesn't seem to be noticeably slower. echo 1 /proc/sys/vm/drop_caches Clears all filesystem caches, so that everything has to be loaded from disk, that's essentially what you have after a cold boot. - Sebastian --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
Re: Review Request 113415: Port KSplashQML to Qt5 + remove XLib stuff + replace x11eventFilter with DBus interface (Wayland++)
On Oct. 24, 2013, 4:17 p.m., Fredrik Höglund wrote: What's the cold startup time like for KSplashQML compared to KSplashX? Let's not forget that the reason KPlash was rewritten to only depend on X + libjpeg + libpng was so that the startup time would be limited by the time it takes to load the images for the theme. Martin Gräßlin wrote: but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. Ivan Čukić wrote: We haven't done any real benchmarks, but it doesn't seem to be noticeably slower. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. Thomas Lübking wrote: Isn't the splash screen supposed to cover (also) those loading costs? If the splashscreen took like 5 seconds to load but then could finish in 500ms, we would simply not require a splash screen in the first place. Martin Klapetek wrote: I've never done any (useful) startup benchmarks, but I'd be happy to try. Can you suggest some ways to measure cold startup time? Then I'll test both and post my findings. Ivan Čukić wrote: If loading Qt took 5 seconds, loading the rest would take 10 minutes. As I said doesn't seem to be noticeably slower. Sebastian Kügler wrote: echo 1 /proc/sys/vm/drop_caches Clears all filesystem caches, so that everything has to be loaded from disk, that's essentially what you have after a cold boot. but that's a long time ago, isn't it? So Moore's Law... Anyway could be interesting to do such comparisons on spinning metal. It was a long time ago, but that doesn't mean you should undo what is essentially a bug fix without checking if the bug is still reproducible. It should not slow down the boot - if it does start slower than ksplashx, it speeds up loading the workspace for the same amount by having Qt preloaded. The point isn't whether it slows down the boot process or not, the point is that the splash screen should appear immediately so the user isn't left staring at a blank screen for N seconds. - Fredrik --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/#review42299 --- On Oct. 24, 2013, 2:19 p.m., Martin Klapetek wrote: --- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/113415/ --- (Updated Oct. 24, 2013, 2:19 p.m.) Review request for kde-workspace and KDE Frameworks. Repository: kde-workspace Description --- Ports KSplashQML to Qt5 and strips any XLib stuff. Martin G. suggested to try if it would simply work without XLib/XCB bits at all and it does work in testing mode, I'm unsure if it also works after actual login, but I don't see why it wouldn't. KSplash was also receiving the stage messages via XAtoms, which I replaced with simple DBus interface. That means that all the parts in the login sequence need to be updated (I'll do it) to emit dbus signal instead of sending X message. I used the same API as the XAtoms were using, but let me know if you think this could be changed. In some future I'd probably also change the stages resolution as it's not too robust and does not help paralelization much. I have a plan in my head but that's for another commit. I'd also like to remove the old KSplashX and port its default theme to QML and provide together with KSplashQML as Classic or KDE 4 theme. Diffs - ksplash/ksplashqml/CMakeLists.txt 8cd5305 ksplash/ksplashqml/SplashApp.h 3c55be9 ksplash/ksplashqml/SplashApp.cpp 7c528d0 ksplash/ksplashqml/SplashWindow.h 9680c1e ksplash/ksplashqml/SplashWindow.cpp 4417643 ksplash/ksplashqml/SystemInfo.h 7a74554 ksplash/ksplashqml/main.cpp c2722ab9 ksplash/ksplashqml/org.kde.KSplash.xml PRE-CREATION ksplash/ksplashqml/themes/Minimalistic/main.qml e4fb8b8 Diff: http://git.reviewboard.kde.org/r/113415/diff/ Testing --- Theme successfully loads, displays fullscreen (Plasma panels are not covered, not sure why), all stages passes, then it terminates itself. Thanks, Martin Klapetek
