QDashStroker::end() in qgl_mac.o
QDashStroker::end() in qwindowsurface_gl.o
QDashStroker::end() in qpixmapdata_gl.o
QDashStroker::end() in qgl.o
QDashStroker::end() in qglpixelbuffer.o
QDashStroker::end() in qglframebufferobject.o
QDashStroker::end() in qpaintengine_opengl.o
QDashStroker::end() in qglpixmapfilter.o
IMHO, this change introduces a new error.
It can successfully delete the mocinclude.tmp file but can’t recreate it.
In building state, mingw32-make will create some thing like “tmpmocdebug_sharedmocinclude.tmp” or “tmpmocrelease_sharedmocinclude.tmp”
in webkit\Webcore directory not the “mocinclude.tmp”
in the expected temp directory.
My suggestion is that the WIN_INCLUDETEMP variable should not be changed.
You should create new variable such as DEL_INCLUDETEMP instead.
ex. DEL_INCLUDETEMP=WIN_INCLUDETEMP DEL_INCLUDETEMP~=s,/,\,g system($$QMAKE_DEL_FILE $$DEL_INCLUDETEMP > NUL 2>&1)
this (almost useless) commit was added to “stage 1 of incorporating Unicode 5.2.0 support” patchset because of any changes to char.h leads to all-things-rebuild which slows down testing of latest changes…
I don’t think you should allow mmaping beyond EOF, but in the bug I have reported (http://bugreports.qt.nokia.com/browse/QTBUG-6330) the problem is that “st.st_size” does not reflect the real EOF.
Hence the warning is a false warning, in this case.
It doesn’t work. When editing “qfontdatabase_win.cpp” file then changing the font of a QDateEdit widget (with enabled calendarPopup and Thai localed settings), and recompiling my project, it still render wrongly. What should I do?
P.S. I designed the UI using Qt Designer integrated via Qt Add-in 1.1.4 for VS 2008 and I installed the Qt SDK 4.6.2 for VS 2008 on Windows 7 Professional (64-bit) machine.
Does this mean the following docs are not correct any more ?
network/socket/qtcpsocket.cpp: around line 61 \bold{Note:} TCP sockets cannot be opened in QIODevice::Unbuffered mode.
corelib/io/qiodevice.cpp: around line 285 ...for example, QTcpSocket does not support \c Unbuffered mode, ...
(and maybe somewhere else)
I've checked that this patch fixes the memory leak that the merge request http://qt.gitorious.org/qt/qt/merge_requests/1129 is also targeting. Here’s an excerpt from a gdb session which shows that the texture cache insertion and removal keys are identical which ensures the cache does not grow needlessly.
Breakpoint 3, QGLTextureCache::remove (this=0x5bb310, key=219043332096) at qgl.cpp:1840
Breakpoint 1, QGLTextureCache::insert (this=0x5bb310, ctx=0x4e40b8, key=219043332096, texture=0x4414e288, cost=212) at qgl.cpp:1833
Breakpoint 3, QGLTextureCache::remove (this=0x5bb310, key=219043332096) at qgl.cpp:1840
Breakpoint 1, QGLTextureCache::insert (this=0x5bb310, ctx=0x4e40b8, key=219043332096, texture=0x44186580, cost=212) at qgl.cpp:1833
I squashed the 3 commits, fixed (I think) indentation (thanks for the link), added Task-number field and tried to explained why I think fuzzy comparison is not a so bad solution in our case in the new commit comment.
Except I'm still not getting a finishing signal emitted.
QProcessManager stays forever waiting for the process to die — thereby never emitting the finished() signal.
I've spent days on this and was super-excited when I saw this patch posted.
but alas still no joy.
I tried uncommenting the defines for QPROCESS_DEBUG in qprocess.cpp and qprocess_unix.cpp but I see nothing at all that gives me hint.
Just I'm just venting here. I'm not sure what information I can give to help debug this.
Allen, I'm sorry but i don’t know what you are talking about. This change has nothing to do with QProcess.
Please comment on the appropriate bug. (or open a new one) (and make sure we have the proper testcase.)
Avoid commenting on gitorious patches. we usually do not read those comment.
Commit d0761cf:
There is include for errno.h in qsystemerror.cpp ifdeffed out of wince just above the bit where errno.h is included in this patch. It should be just moved outside the ifdefs if it is needed on all builds.
#if !defined(Q_OS_WINCE)
# include
# if defined(Q_CC_MSVC)
# include
# endif
#else
# include
#endif
Commit 11b3fc4:
Remove comment "# replace with actual SDK name" strings from the two new qmake.conf files.
Commit fd516a4:
Has unrelated whitespace addition on line 34 of gzguts.h
Commit 7df17d7:
1) Has unrelated whitespace removal, line 100 in qfunctions_wince.h
2) Two identical "#ifdef __cplusplus" scopes without anything else defined in-between. Could be combined.
Commit ae3aae2:
Two identical "#ifndef SQLITE_OMIT_LOCALTIME" scopes without anything else defined in-between. Could be combined.
Commit 085e793:
1) examples.pro: Why is gestures subdir removed from all platforms?
2) network.pro: Why is download subdir removed from all platforms?
3) demos/wincepkgrules.pri: Contents of this file should be similar to examples/wincepkgrules.pri.
4) examples/wincepkgrules.pri: RC_FILE points under demos dir. I don't think there should be dependencies to demos in examples.
Question related to lines 240 and 241: Is the syntax supposed to be “user@domain” or “domain\user”, so in fact the line 240 should be d->userDomain = user.mid(separatorPosn); and not “user.left(separatorPosn)”.
Loosely related:
I've been looking at this code because I'm still having trouble with Qt bug 15472, which I got working in Qt 4.7.1 by patch. I am now trying to apply the patch (in above mentioned bug) to 4.7.4 unsuccesfully. I don’t know why I'm having this issue, it seems as if it was corrected alread (the NTLM v2 issue).
Could this be the reason for QtOpenGL not compiling on a Mac (using Frameworks)?
g++ -headerpad_max_install_names -Xarch_i386 -mmacosx-version-min=10.4 -arch i386 -single_module -dynamiclib -compatibility_version 4.6 -current_version 4.6.0 -install_name /usr/local/Trolltech/Qt-4.6.0/lib/QtOpenGL.framework/Versions/4/QtOpenGL -o QtOpenGL .obj/release-shared/qgl_mac.o .obj/release-shared/qglpixelbuffer_mac.o .obj/release-shared/qgraphicssystem_gl.o .obj/release-shared/qwindowsurface_gl.o .obj/release-shared/qpixmapdata_gl.o .obj/release-shared/qgl.o .obj/release-shared/qglcolormap.o .obj/release-shared/qglpixelbuffer.o .obj/release-shared/qglframebufferobject.o .obj/release-shared/qglextensions.o .obj/release-shared/qpaintengine_opengl.o .obj/release-shared/qglpixmapfilter.o .obj/release-shared/moc_qgl.o .obj/release-shared/moc_qgl_p.o -F/Users/njeiseck/projects/qt/qt-clone/lib -L/Users/njeiseck/projects/qt/qt-clone/lib -framework OpenGL -framework AGL -framework QtGui -L/tmp/qt-stuff/source/qt-mac-cocoa-commercial-src-4.4.1/lib -F/tmp/qt-stuff/source/qt-mac-cocoa-commercial-src-4.4.1/lib -framework Carbon -framework AppKit -framework QtCore -lz -lm -framework ApplicationServices
Undefined symbols:
“QStrokerOps::end()”, referenced from:
…
Fixed. Change will appear later today.
See commit comment.
yay! eventually…
That is bad look:
http://lit-uriy.narod.ru/temp/qt_4.6_doc_2.png
bad include in line 44 @ src/gui/styles/qproxystyle.cpp
please replace ‘null EventFilter’ by ‘0’
wrong task number :)
That should be
hasMoreLengthVariants
not
hasMoreLenghtVariants
This looks like it may be related to tracker issue 258216, not certain though.
IMHO, this change introduces a new error.
It can successfully delete the mocinclude.tmp file but can’t recreate it.
In building state, mingw32-make will create some thing like “tmpmocdebug_sharedmocinclude.tmp” or “tmpmocrelease_sharedmocinclude.tmp”
in webkit\Webcore directory not the “mocinclude.tmp”
in the expected temp directory.
My suggestion is that the WIN_INCLUDETEMP variable should not be changed.
You should create new variable such as DEL_INCLUDETEMP instead.
ex.
DEL_INCLUDETEMP=WIN_INCLUDETEMPDEL_INCLUDETEMP~=s,/,\,gsystem($$QMAKE_DEL_FILE $$DEL_INCLUDETEMP > NUL 2>&1)Neat!
this (almost useless) commit was added to “stage 1 of incorporating Unicode 5.2.0 support” patchset because of any changes to char.h leads to all-things-rebuild which slows down testing of latest changes…
I don’t think you should allow mmaping beyond EOF, but in the bug I have reported (http://bugreports.qt.nokia.com/browse/QTBUG-6330) the problem is that “st.st_size” does not reflect the real EOF.
Hence the warning is a false warning, in this case.
Merged in 4.7
QString class has missing overload method
QString::number(float, char f=‘g’, int pred=6)
Added update on tst_qurl.cpp separately.
Erased binary file that had been committed together with the other files.
Erased binary file that had been committed together with the other files.
Hi
there is a problem with this line:
\"$$system(readelf -l /bin/ls | perl -n -e '$$prog')\":
ELF_INTERPRETER will set be to : \“/lib/ld-linux.so.2\”
in src/corelib/Makefile you will gonna have:
-DELF_INTERPRETER=\“/lib/ld-linux.so.2\”
and eventually a compile error in qlibraryinfo.cpp:491 in src/corelib/global
I worked around this by hard-coding “/lib/ld-linux.so.2” in this file instead of using the value of ELF_INTERPRETER. (like it was before as I can see)
I am for sure not the first one trying to configure and compile Qt 4.6.2 ;) so I wonder how it compiles with this issue ? Am I missing something ?
Regards.
Squashed previous steps into this one change.
It doesn’t work. When editing “qfontdatabase_win.cpp” file then changing the font of a QDateEdit widget (with enabled calendarPopup and Thai localed settings), and recompiling my project, it still render wrongly. What should I do?
Thank you very much.
P.S. I designed the UI using Qt Designer integrated via Qt Add-in 1.1.4 for VS 2008 and I installed the Qt SDK 4.6.2 for VS 2008 on Windows 7 Professional (64-bit) machine.
Should fix QTBUG-186 and QTBUG-5066
QLocale time to string conversion localization using QLocale::amText() and QLocale::pmText() added.
QLocale::(am|pm)Text() never returns an empty strings
I think, the fix will now cause out-of-bounds write if the new base_pixels has less memory allocated than the original pixels.
Anyway, I am very sceptic about copying pixel data in cases of resizing. Maybe the memcpy should only happen when the sizes are not changing?
Also, what if the pixels have already been registered with the cg_data?
And, I think, there is a “delete pixels” missing before line 641.
At least, we are observing a memory leak here with QPixmap::loadFromData() on Mac.
Filed a bug report for this issue:
http://bugreports.qt.nokia.com/browse/QTBUG-12873
Sorry, should have been “free(pixels)” instead of “delete pixels”.
I updated it to do it more properly with -webkit-debug. I also added a separate commit to the merge request for the same fix on Windows.
This makes build on Windows x64 fail (mingw-w64):
This should be reverted…
Something wrong happened with the commit selection, trying again…
Change long tmp to quint64 tmp to fix this.
Change to a reference.
This shouldn’t be included on SSE4.1 machines.
http://bugreports.qt.nokia.com/browse/QTBUG-13623
Changed patches
Does this mean the following docs are not correct any more ?
network/socket/qtcpsocket.cpp: around line 61
\bold{Note:} TCP sockets cannot be opened in QIODevice::Unbuffered mode.corelib/io/qiodevice.cpp: around line 285
...for example, QTcpSocket does not support \c Unbuffered mode, ...(and maybe somewhere else)
I've checked that this patch fixes the memory leak that the merge request http://qt.gitorious.org/qt/qt/merge_requests/1129 is also targeting. Here’s an excerpt from a gdb session which shows that the texture cache insertion and removal keys are identical which ensures the cache does not grow needlessly.
Breakpoint 3, QGLTextureCache::remove (this=0x5bb310, key=219043332096) at qgl.cpp:1840
Breakpoint 1, QGLTextureCache::insert (this=0x5bb310, ctx=0x4e40b8, key=219043332096, texture=0x4414e288, cost=212) at qgl.cpp:1833
Breakpoint 3, QGLTextureCache::remove (this=0x5bb310, key=219043332096) at qgl.cpp:1840
Breakpoint 1, QGLTextureCache::insert (this=0x5bb310, ctx=0x4e40b8, key=219043332096, texture=0x44186580, cost=212) at qgl.cpp:1833
Thank you for having “pre-reviewed” my commit.
I squashed the 3 commits, fixed (I think) indentation (thanks for the link), added Task-number field and tried to explained why I think fuzzy comparison is not a so bad solution in our case in the new commit comment.
This patch does help me out — somewhat.
Except I'm still not getting a finishing signal emitted.
QProcessManager stays forever waiting for the process to die — thereby never emitting the finished() signal.
I've spent days on this and was super-excited when I saw this patch posted.
but alas still no joy.
I tried uncommenting the defines for QPROCESS_DEBUG in qprocess.cpp and qprocess_unix.cpp but I see nothing at all that gives me hint.
Just I'm just venting here. I'm not sure what information I can give to help debug this.
-Allen
oh, I forgot to mention…
before I applied this patch I was seeing a message
“WARNING: QObject::startTimer: timers cannot be started from another thread”
that message seems to be gone now. but something else must be buggy with QProcessManager or QProcess. Which I knew what.
Allen, I'm sorry but i don’t know what you are talking about. This change has nothing to do with QProcess.
Please comment on the appropriate bug. (or open a new one) (and make sure we have the proper testcase.)
Avoid commenting on gitorious patches. we usually do not read those comment.
Yes, I was talking complete nonsense. Forget I said anything.
-Allen
Fix the QTBUG-11982: QFile::copy() return false but error() is NoError.
In QFile::copy function after call serError() it will call colse(), then here will unset the error.
Task number: QTBUG-11982
Commit d0761cf:
Commit 11b3fc4: Remove comment "# replace with actual SDK name" strings from the two new qmake.conf files. Commit fd516a4: Has unrelated whitespace addition on line 34 of gzguts.h Commit 7df17d7: 1) Has unrelated whitespace removal, line 100 in qfunctions_wince.h 2) Two identical "#ifdef __cplusplus" scopes without anything else defined in-between. Could be combined. Commit ae3aae2: Two identical "#ifndef SQLITE_OMIT_LOCALTIME" scopes without anything else defined in-between. Could be combined. Commit 085e793: 1) examples.pro: Why is gestures subdir removed from all platforms? 2) network.pro: Why is download subdir removed from all platforms? 3) demos/wincepkgrules.pri: Contents of this file should be similar to examples/wincepkgrules.pri. 4) examples/wincepkgrules.pri: RC_FILE points under demos dir. I don't think there should be dependencies to demos in examples.There is include for errno.h in qsystemerror.cpp ifdeffed out of wince just above the bit where errno.h is included in this patch. It should be just moved outside the ifdefs if it is needed on all builds.
Question related to lines 240 and 241: Is the syntax supposed to be “user@domain” or “domain\user”, so in fact the line 240 should be d->userDomain = user.mid(separatorPosn); and not “user.left(separatorPosn)”.
Loosely related:
I've been looking at this code because I'm still having trouble with Qt bug 15472, which I got working in Qt 4.7.1 by patch. I am now trying to apply the patch (in above mentioned bug) to 4.7.4 unsuccesfully. I don’t know why I'm having this issue, it seems as if it was corrected alread (the NTLM v2 issue).