Change in kio[master]: WIP! Adding SlaveBase newEntry and insertEntryProperty for b...
Mark Gaiser has uploaded a new change for review.
https://gerrit.vesnicky.cesnet.cz/r/348
Change subject: WIP! Adding SlaveBase newEntry and insertEntryProperty for big
speed improvement in listDir calls.
..
WIP! Adding SlaveBase newEntry and insertEntryProperty for big speed
improvement in listDir calls.
WIP because the tests:
- copyjob
- testtrash
still fail.. I don't know why yet.
This commit adds adds the following functions:
- newEntry()
- insertEntryProperty(uint field, const QString value)
- insertEntryProperty(uint field, const long long value)
The existing listEntry function (that was introduced in KIO 5.0) is now
deprecated. Internally it is
now using the new functions.
These methods are introduced to completely phase out a need for UDSEntry on the
slave side.
Slaves can now - with those function - directly insert stat data in a buffer.
This is much faster then going through QDataStream which does all kinds of
magic to be cross platform compatible.
Since QDataStream is dropped, the byte order management (that it did to be
cross platform usable) is obviously
gone as well. It shouldn't be an issue since KIO isn't used over the network.
It's all living on the
same machine thus the same byte order anyway.
Some numbers for performance comparision.
Running KIO::listDir on a folder with 500.000 files.
Before this patch: ~1300ms
After this patch: ~800ms
Native speed (without KIO, just C++ and storing the stat data in a vector):
~300ms
Change-Id: Ibb61f0c3a406337c54d066cac169ca534ec1d8b2
CHANGELOG: Introduce newEntry and insertEntryProperty for better listdir
performance. Deprecate listEntry.
---
M autotests/udsentrytest.cpp
M src/core/slavebase.cpp
M src/core/slavebase.h
M src/core/udsentry.cpp
M src/core/udsentry.h
M src/ioslaves/file/file.cpp
M src/ioslaves/file/file.h
M src/ioslaves/file/file_unix.cpp
8 files changed, 267 insertions(+), 100 deletions(-)
git pull ssh://gerrit.vesnicky.cesnet.cz:29418/kio refs/changes/48/348/1
diff --git a/autotests/udsentrytest.cpp b/autotests/udsentrytest.cpp
index 04e072e..066fd86 100644
--- a/autotests/udsentrytest.cpp
+++ b/autotests/udsentrytest.cpp
@@ -183,17 +183,21 @@
{
QDataStream stream(data, QIODevice::WriteOnly);
foreach (const QVectorUDSTestField testCase, testCases) {
-stream testCase.count();
+const int size = testCase.count();
+stream.writeRawData(reinterpret_castconst char*(size),
sizeof(int));
foreach (const UDSTestField field, testCase) {
uint uds = field.m_uds;
-stream uds;
+stream.writeRawData(reinterpret_castconst char*(uds),
sizeof(uint));
if (uds KIO::UDSEntry::UDS_STRING) {
-stream field.m_string;
+const QString str = field.m_string;
+int size = str.size();
+stream.writeRawData(reinterpret_castconst char*(size),
sizeof(int));
+stream.writeRawData(reinterpret_castconst
char*(str.utf16()), sizeof(ushort) * size);
} else {
Q_ASSERT(uds KIO::UDSEntry::UDS_NUMBER);
-stream field.m_long;
+stream.writeRawData(reinterpret_castconst
char*(field.m_long), sizeof(long long));
}
}
}
diff --git a/src/core/slavebase.cpp b/src/core/slavebase.cpp
index db1c2ee..2e0333c 100644
--- a/src/core/slavebase.cpp
+++ b/src/core/slavebase.cpp
@@ -37,7 +37,6 @@
#include QtCore/QDateTime
#include QtCore/QElapsedTimer
#include QtCore/QCoreApplication
-#include QtCore/QDataStream
#include kconfig.h
#include kconfiggroup.h
@@ -74,6 +73,59 @@
namespace KIO
{
+// This class is a small QByteArray wrapper which provides some convenience
functions.
+class EntryBuffer
+{
+public:
+EntryBuffer()
+: m_buffer()
+, m_pos(0)
+{
+// Initialize m_buffer with 8KB. That should be enough room for small
folders (lets say ~ 70 entries).
+m_buffer.resize(8192);
+}
+
+// Return the raw QByteArray buffer. The user should only use this
function to get a buffer of the written bytes.
+// That would be a call like: buffer().left(pos()).
+QByteArray buffer()
+{
+return m_buffer;
+}
+
+int pos()
+{
+return m_pos;
+}
+
+void seek(int pos)
+{
+m_pos = pos;
+}
+
+void reset()
+{
+m_pos = 0;
+}
+
+// This function writes the provided data in the internam m_buffer. The
user should not write more then 4096 bytes at once!
+qint64 write(const char *data, qint64 len)
+{
+// Increment the m_buffer in chunks of 4KB.
+// Realistically this should only happen for larger folders (70
entries).
+if ((m_pos + len) m_buffer.size()) {
+m_buffer.resize(m_buffer.size() +
Re: Review Request 122330: Associate *.qmltypes and *.qmlproject files with the text/x-qml mime type
--- This is an automatically generated e-mail. To reply, visit: https://git.reviewboard.kde.org/r/122330/ --- (Updated Jan. 31, 2015, 6:48 p.m.) Review request for KDE Frameworks, kdelibs and KDevelop. Repository: kcoreaddons Description --- The KDevelop QML/JS plugin sometimes needs to parse .qmltypes files (in order to list the content of installed QML modules, for instance), but KDevelop requires that files parsed using the QML/JS plugin have the text/x-qml mime type. Because .qmltypes and .qmlproject files are valid QML files (they follow the standard QML syntax), this patch proposes to add these extensions to the ones associated with the text/x-qml mime type. Diffs - src/mimetypes/kde5.xml cc9f71e Diff: https://git.reviewboard.kde.org/r/122330/diff/ Testing --- Installing kcoreaddons and updating the system MIME database allows KDevelop to detect that files with the .qmltypes and .qmlproject extensions have to be parsed using the QML/JS language support plugin. This allows QML files to use QML modules installed system-wide (and fixes the unit tests of kdev-qmljs). Thanks, Denis Steckelmacher ___ Kde-frameworks-devel mailing list Kde-frameworks-devel@kde.org https://mail.kde.org/mailman/listinfo/kde-frameworks-devel
Re: Build failed in Jenkins: kde-baseapps_master_qt5 #213
On Wed, Jan 28, 2015 at 4:24 PM, David Faure fa...@kde.org wrote: On Wednesday 28 January 2015 14:41:04 KDE CI System wrote: [ 74%] Building CXX object dolphin/src/CMakeFiles/kcm_dolphinnavigation.dir/dolphin_generalsettings.cp p.o http://build.kde.org/job/kde-baseapps_master_qt5/ws/dolphin/src/search/dol phinsearchbox.cpp:42:44: fatal error: Baloo/NaturalFileQueryParser: No such file or directory #include Baloo/NaturalFileQueryParser ^ compilation terminated. Hi Vishesh, Any idea what changed in baloo that causes this CI error? Missing dependency maybe? Yes. I moved the file from baloo to baloo-widgets. Let me try to fix it. -- Vishesh Handa ___ Kde-frameworks-devel mailing list Kde-frameworks-devel@kde.org https://mail.kde.org/mailman/listinfo/kde-frameworks-devel
Build failed in Jenkins: kde-baseapps_master_qt5 #218
See http://build.kde.org/job/kde-baseapps_master_qt5/218/changes Changes: [emmanuelpescosta099] Port away from KGlobalSettings::naturalSorting() by moving it to Dolphin's GeneralSettings -- [...truncated 3048 lines...] http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:489:20: warning: ‘bool QTest::kWaitForSignal(QObject*, const char*, int)’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/qtest_kde.h:48) [-Wdeprecated-declarations] QVERIFY(QTest::kWaitForSignal(view, SIGNAL(viewCompleted(KonqView*)), 1)); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:489:80: warning: ‘bool QTest::kWaitForSignal(QObject*, const char*, int)’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/qtest_kde.h:48) [-Wdeprecated-declarations] QVERIFY(QTest::kWaitForSignal(view, SIGNAL(viewCompleted(KonqView*)), 1)); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp: In member function ‘void ViewMgrTest::testAddTabs()’: http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:510:17: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:510:60: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp: In member function ‘void ViewMgrTest::testSaveProfile()’: http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:704:19: warning: ‘KUrl’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/kurl.h:109) [-Wdeprecated-declarations] const KUrl url(data:text/html, pHello World/p); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:708:20: warning: ‘KUrl’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/kurl.h:109) [-Wdeprecated-declarations] const KUrl url2(data:text/html, pview2/p); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:710:17: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:710:60: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp: In member function ‘void ViewMgrTest::testCloseOtherTabs()’: http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:762:17: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/konqueror/src/tests/konqviewmgrtest.cpp:762:60: warning: ‘KTabWidget’ is deprecated (declared at /srv/jenkins/install/linux/x86_64/g++/kf5-qt5/frameworks/kdelibs4support/inst/include/KF5/KDELibs4Support/ktabwidget.h:44) [-Wdeprecated-declarations] KTabWidget* tabWidget = mainWindow.findChildKTabWidget*(); ^ http://build.kde.org/job/kde-baseapps_master_qt5/ws/build/dolphin/src/dolphin_searchsettings.h:0: Note: No relevant classes found. No
Re: [Kde-games-devel] Freeze in 6 weeks
Albert, Frameworks developers/maintainers, On Thu, Jan 29, 2015 at 12:43 PM, Albert Astals Cid aa...@kde.org wrote: El Dijous, 29 de gener de 2015, a les 10:05:01, Inge Wallin va escriure: On Wednesday, January 28, 2015 20:47:44 Albert Astals Cid wrote: El Dimecres, 28 de gener de 2015, a les 14:04:14, Inge Wallin va escriure: On Wednesday, January 28, 2015 07:38:23 laurent Montel wrote: I am agree with Albert if there is a problem with network-transparency on MacOsX it's better to fix it that remove it. That is not the issue. The issue is that it was promised that if you developed your program with KDE Frameworks 5 instead of KDElibs, you would get a more lightweight program that did not need any daemons to run. Now this turns out to be not true. *That* is what we want, not to remove network transparency per se. But if the promise from frameworks 5 is not kept, what is there to be done? Really? Who promised you that? That's never been the promise, the promise of KF5 is that it's more modular, and yes there's some libs that don't need extra daemons and some that do need them. I actually tried to do some research to find this promise in writing but except for pointing it out as a problem in an article on the dot I couldn't find any written promise. The list on http://api.kde.org/frameworks-api/frameworks5-apidocs/ has three items, functional, solution, integration. As far as i remember, functional means you can use it standalone, solution means it needs a daemon and integraion means it needs more stuff (more daemons? Is this categorization correct? Has it been checked with a dependency analysis to see that each functional framework is usable standalone? I understood the tiers concept but never did get my head around the functional/integration/solution concept yet, but would like to. At any rate, I think KNewStuff which I'm very familiar with doesn't seem to be usable standalone as the functional status it has implies. It depends on and uses KIO for all network downloads and uploads, which requires a daemon as far as I can tell (and is a solution) so should KNewStuff be changed to a solution also probably? But I do remember bringing up this problem in a number of conversations with frameworks developers early in the development cycle. One of those times was in a frameworks meeting at Akademy in Tampere. And the developers did say that bringing down the amount of infrastructure that the first KDE application on an alien platform started was one of the motivations for frameworks. Not the only one, of course, but still an important one. And this problem will be in other application. It's not just a problem with kdegames so fix it for kdegames will fix for all application. We can't remove feature each time we need to fix on a specific platform no ? Yes, it will be in other applications. But that is the price you have to pay for platform portability sometimes. I never thought it was reasonable to have to start the full KDE desktop infrastructure just to run an application from the KDE on, say, Gnome. But that's what we had to do back in the kdelibs days. Frameworks 5 was supposed to get rid of that but now it turns out that it doesn't. The question is: is this just not yet implemented or was this design goal abandoned? What is the full KDE desktop infraestructure for you? Let's skip the word full, since I don't really know every nook and crannie of the kde infrastructure. But to start several daemons to be able to access a file on another server seems like overkill. Even one, in fact. Why starting a program to do something is an overkill? If i want to convert a pdf to a ps, running pdftops seems totally legit to me. Cheers, Albert I understand (or at least I think I do) that kded is an optimisation that loads all the shared libraries once so that all the subsequent kde applications don't have to do it. Instead they are forked off from kded with the libraries already loaded. This is a good idea if you know for sure that you will be running a whole slew of such applications. But this is not a very strong guess on non-KDE platforms and even less so on non-linux platforms. -Inge Cheers, Albert This is what we need to focus on. Network transparency is *not* the issue. Cheers, Albert Cheers, Ian W. ___ kde-games-devel mailing list kde-games-de...@kde.org https://mail.kde.org/mailman/listinfo/kde-games-devel ___ kde-games-devel mailing list kde-games-de...@kde.org https://mail.kde.org/mailman/listinfo/kde-games-devel
Change in kio[master]: UDSEntry documentation fix.
Mark Gaiser has uploaded a new change for review. https://gerrit.vesnicky.cesnet.cz/r/349 Change subject: UDSEntry documentation fix. .. UDSEntry documentation fix. Change-Id: I57151f76bf086b7d8108391c97dbdaf548e41251 --- M src/core/udsentry.h 1 file changed, 3 insertions(+), 3 deletions(-) git pull ssh://gerrit.vesnicky.cesnet.cz:29418/kio refs/changes/49/349/1 diff --git a/src/core/udsentry.h b/src/core/udsentry.h index 13d50ea..e42ef23 100644 --- a/src/core/udsentry.h +++ b/src/core/udsentry.h @@ -100,14 +100,14 @@ void reserve(int size); /** - * insert field with numeric value + * insert field with string value * @param field numeric field id - * @param value + * @param value to set */ void insert(uint field, const QString value); /** - * insert field with string value + * insert field with numeric value * @param field numeric tield id * @param l value to set */ -- To view, visit https://gerrit.vesnicky.cesnet.cz/r/349 To unsubscribe, visit https://gerrit.vesnicky.cesnet.cz/r/settings Gerrit-MessageType: newchange Gerrit-Change-Id: I57151f76bf086b7d8108391c97dbdaf548e41251 Gerrit-PatchSet: 1 Gerrit-Project: kio Gerrit-Branch: master Gerrit-Owner: Mark Gaiser mark...@gmail.com Gerrit-Reviewer: David Faure fa...@kde.org ___ Kde-frameworks-devel mailing list Kde-frameworks-devel@kde.org https://mail.kde.org/mailman/listinfo/kde-frameworks-devel
Jenkins build is back to stable : plasma-framework_master_qt5 » All,LINBUILDER #984
See http://build.kde.org/job/plasma-framework_master_qt5/Variation=All,label=LINBUILDER/984/changes ___ Kde-frameworks-devel mailing list Kde-frameworks-devel@kde.org https://mail.kde.org/mailman/listinfo/kde-frameworks-devel
