* contrib/hbqt/hbmk2_qt.hb
* contrib/hbqt/qtcore/hbqt.h
* contrib/hbqt/qtcore/hbqt_init.cpp
- Removed: a hack introduced some months back where
C++ static items were held in a list them released
at exit of an application. This was double work and
now is not needed at all.
* contrib/hbqt/qtcore/hbqt_bind.cpp
- Reverted: last fix where <qtObject> was compared against
a numeric, which, BTW was a stupid implementation.
A pointer casted to a numeric is always true.
* contrib/hbqt/qtgui/hbqt_init.cpp
- Deleted: deleting the QApplication object.
It was breaking such code:
oWnd:connect( QEvent_Close, {|| QApplication():quit() } )
QApplication():exec()
the reason was any code executed after
QApplication():quit() was rendered unexcutable because
appln was forced to stop execution.
* contrib/hbqt/qtcore/hbqt_hbqevents.cpp
+ Reintroduced: QEvent_Close's return value to true.
It is necessary to intercept this event at appln level.
* contrib/hbqt/qtcore/qth/QLibraryInfo.qth
* contrib/hbqt/qtgui/qth/QDesktopServices.qth
+ Reworked: constructors.
Previous implementation was wrong.
Applied Francesco's thought and it worked, thanks.
* contrib/hbqt/qtgui/qth/QSound.qth
+ Added: two slot methods.
* utils/hbmk2/hbmk2.hbp
* utils/hbmk2/hbmk2.prg
* utils/hbmk2/Makefile
* minor changes to do not require std.ch
+ enabled -u option for hbmk2 source code.
* contrib/hbide/idedocks.prg
! Fixed: one more QIcon() regression, though not used.
* contrib/hbqt/qtcore/hbqt_hbqslots.cpp
! Fixed: to pass correct class name for some type of slots
<pList> in constructors.
* contrib/hbqt/qtcore/hbqt_bind.cpp
* contrib/hbqt/qtcore/qth/QLibraryInfo.qth
* contrib/hbqt/qtgui/qth/QDesktopServices.qth
! Implemented: mechanism to generate a Harbour level class
without the need of a Qt level object for those classes
where methods are executed by Qt code only.
This fixes the bug issue reported by Ligui.
* contrib/hbqt/qtcore/hbqt_bind.cpp
* contrib/hbqt/qtcore/hbqt_misc.prg
+ Implemented: DESTRUCTOR FUNCTION <somePrgOrCFunc()>
* contrib/hbqt/qtgui/hbqt_init.cpp
* contrib/hbqt/qtcore/hbqt_init.cpp
! Fixed: some rare signals issued by Qt with NULL pointers.
This fixes regression reported on user-list a few days back.
And probably fixes many-many unexplained GPF's.
* utils/hbmk2/hbmk2.prg
* marked the file as UTF8
* added accent to my own name. let's see on which platforms
and configs it appears alright (second "a" should appear
with an acute accent: "a'")
[ms-dos and OS/2 is not expected to work]
* INSTALL
% tuned to not require regular updates when new versions of
certain linked products are release (ie. deleted latest
version specific links and references)
* harbour/include/hbclass.ch
+ added support for
DESTRUCTOR FUNCTION <FuncName>[()]
in class declaration. It allows to define as object
destructors external functions, i.e. defined in C
* harbour/src/rtl/valtype.c
* harbour/include/harbour.hbx
+ added new PRG function HB_ISEVALITEM()
It returns true if item can be evaluated by EVAL() function,
i.e. for codeblocks and function symbols.
In the future it can be extended to accept objects with :EVAL()
methods.
* contrib/hbqt/qtcore/hbqt_bind.cpp
+ Implemented: Events, Slots, Destroyers on thread level.
Previously those were created per object level. This greatly
decreased the os-resource count and also simplifies the code.
Thanks to Przemek for the tip.
+ Implemented: release of all objects per thread level.
hbQT employs internal counter to identify threads.
All widgets are created in one global list which has
made it possible to use Qt's blocking widgets, like
QMessageBox, in Harbour MT applications. It seems that
Qt looks for the event processing only in main appln
thread and hence these widgets were not usable in
other threads. The testbed is hbqt/tests/demoqt.prg
which is now multithreaded. You check it with
menu option <Dialog><Another Dialog>.
+ Formatting: in tracelog entries.
* contrib/hbqt/qtcore/hbqt_hbqevents.cpp
- Commented out: a hack where QClosEvent was being treated
differently then other events.
* contrib/hbqt/tests/demoqt.prg
* Parented: QMessageBox() ( Read below )
* contrib/hbide/idedocks.prg
- Removed: one hack which at begining was there to activate
QMainWindow's docking area resizable.
* contrib/hbide/idemain.prg
+ Changed: ::cWrkCodec to be "EN" by default.
It fixes recent regression in hbIDE for certain users.
* contrib/hbide/idemisc.prg
* Parented: QMessageBox() should always be constructed with a
parent because of special handelling of blocking dialog
where parent is treated slightly different than other
Qt widgets. This fixes a long-standing bug in hbIDE where
appnl was rendered in indefinite loop when "X" button
was employed to terminate and confirm dialog would
not be terminated internally.
So for QMessageBox() and, to be on the safer side, all
widets derived from QDialog() should be constructed
with a parent and at the end be reparented with a
just-to-be-deleted-QWidget(), like:
oMB := QMessageBox( oParent )
...
oMB:setParent( QWidget() )
RETURN something
* harbour/src/common/hbffind.c
* ulAttr -> nAttr
+ added support for softlinks in * nix systems.
Now they are reported with "L" attribute by directory() function.
This modification also fixes directory scanning interrupted by
dummy softlinks.
* harbour/contrib/xhb/dirrec.prg
! fixed double added extension to file mask
* contrib/hbqt/qtcore/hbqt_bind.cpp
+ Added: static HB_CRITICAL_NEW( s_qtMtx ); and parts.
+ Tuned: compilable with or without TSD protocol.
To turn on TSD protocol, change
__HBQT_WITH_MT_SUPPORT__NO => __HBQT_WITH_MT_SUPPORT__
* utils/hbmk2/hbmk2.prg
- do not include Harbour bin dir in linker libpath on non-*nix
systems. In could cause some compilers (f.e. mingw) to link
to .dlls stored there (f.e. 3rd party dynamic dlls enabled
with HB_INSTALL_3RDDYN), not letting user supplied libpaths
to kick in. If you for some reason want to include Harbour
bin dir in libpath, use: -L${HB_LIB} option, or to include
dynamic lib dir (emulating old behavior on non-*nix), use
-L${HB_DYN}
! fixed -env option processing after 2012-06-14 12:01 UTC+0200
* INSTALL
+ use win-make in primary build example except mingw32-make,
change wording a little
% tweaked information about large attachment in HOW TO PARTICIPATE
section
* harbour/src/3rd/pcre/Makefile
* disabled JIT support in default builds.
It fixes build process for platforms which do not support MT mode
like some old Linux distors. I also made runtime tests with some
simple expressions and on x86 based CPUs the difference is really
minor. PCRE compiled with JIT is a little bit faster in regex
compilation (~5%) but slower in regex evaluation (10%).
AFAIR completely different results then suggested by documentation.
For tests I used rather simple expression on small data set:
cData := hb_TSToStr( hb_datetime() )
cExp := "[:]([a-zA-Z]|[0-9])[0-9][:]..[.]"
But such short expressions and data are quite common in real
programs.
If someone wants to enable JIT in his build then he can make it
using SUPPORT_JIT macro, i.e.:
export HB_USER_CFLAGS=-DSUPPORT_JIT
* utils/hbmk2/hbmk2.prg
* made --hbinfo to return information only for the top project
(not nested ones)
This should fix regression of not generating .hbx files for
the hbqt special case since 2012-06-06 16:02 UTC+0200
+ added --hbinfo=nested to return information also on nested
projects
+ --hbinfo will now append an 0x0A byte after each JSON
block, so multiple JSON blocks can be parsed by the caller
* contrib/hbqt/hbqt_commons.hbm
* changed to be friendly with '--hbinfo=nested' hbmk2 option
(not used in Harbour make process)
* contrib/hbamf/amfenc.c
* contrib/hbamf/hbamf.hbx
* contrib/hbamf/hbamfobj.prg
* contrib/hbamf/readme.txt
+ applied Aleksander Czajczynski's patch for a missing class.
Thank you very much, I didn't have to touch it, only
autoupdate the .hbx file.
* contrib/hbrun/hbrun.hbp
* utils/hbmk2/hbmk2.hbp
* utils/hbmk2/Makefile
* package/mpkg_win.nsi
* package/winuni/mpkg_win_uni.nsi
; trying to crawl out from this pool of dynamic mud.
* enabled all embedded headers for both hbmk2 and hbrun
so we're back to square one with this, except that
now hbmk2 also holds the full set of core headers not
just a minimal selection. The reason is to keep hbmk2
as the distributable runner tool. It also syncs all
C compilers and platforms.
* hbmk2 built in -static mode again. It will fix the
'make clean' problem, it will make it distributable.
* hbrun changed to be built in -shared mode to allow
to load dynamic modules.
; so this is the state right now (if I didn't make any mistake):
hbmk2:
all embedded core headers
no contrib modules
static build
no dynamic modules
fully movable/distributable as a runner
cannot be self-registered as .hb runner on Windows
hbrun:
all embedded core headers
some contrib modules
shared build when HB_BUILD_CONTRIB_DYN enabled
dynamic modules possible
less movable/distributable, needs harbour dll, dynamic modules need Harbour dir layout
can be self-registered as .hb runner on Windows
; "Ext:" banner on interactive shell will have blue color
if dynamic modules are enabled and grey if not.
* utils/hbmk2/hbmk2.prg
* minor change to show the actual name of the tool
in the error message requesting a -shared build.
* contrib/rddads/ads1.c
* contrib/rddads/adsfunc.c
* contrib/rddads/adsx.c
* src/common/hbfsapi.c
! deleted explicit newlines at EOL in HB_TRACE() calls
* INSTALL
+ added new '0. GUARANTEES AND LIABILITY' section
! updated instructions to create debug build
* utils/hbmk2/hbmk2.prg
* minor cleanups
* utils/hbmk2/Makefile
+ enabled dynamic version of hbmk2 only for mingw builds.
Rest of the targets may or may not work due to CRTL
usage (fprintf()) from HB_COMPILE*() calls which causes
GPF in currently generated shared builds, most probably
due to double copy of statically linked CRTL code.
mingw uses MSVCRT.DLL, so it's not prone to this.
Solution is either to avoid file handling CRTL calls
in compiler code, or to tweak -shared build setting to
death to make it work for all non-mingw compilers.
Both of these is very difficult.
This means that dynamic module loading from scripts
will only work in hbmk2 built with mingw.
; TOFIX: 'make clean' on contribs doesn't work with mingw,
because harbour dll is deleted before hbmk2 is
called, so it cannot be started for cleanup.
* contrib/hbtip/encoder.prg
* contrib/hbtip/encqp.prg
% use hb_default()
* contrib/hbtip/encqp.prg
! fixed RTE when using TIPEncoderQP class from a dynamically loaded hbtip.
It has been reported long time ago, so now I hacked a local
solution which simply removes inherition from this class,
which works around the problem. Hopefully the core can get
the fixes to allow derived classes to be used from dynamic libs
and this can be reverted. In this case only one variable was
inherited, so it was easy to do. BTW I was trying to use hb_MailAssemble()
from a script which triggered this.
TOFIX: Self contained example (tested on win/mingw):
--- hbmk2 test.hb
#require "hbtip"
PROCEDURE Main()
TIPEncoderBase64():New()
--- ->
Error BASE/3003 Cannot find super class 'TIPENCODER': __CLSINSTSUPER
Called from __CLSINSTSUPER(0)
Called from HBCLASS:CREATE(0)
Called from TIPENCODERBASE64(0)
---
* contrib/hbtip/encqp.prg
* contrib/hbtip/hbtip.hbx
+ added two new functions:
TIP_QPENCODE(), TIP_QPDECODE()
formerly only accessible via OOP code
* utils/hbmk2/hbmk2.prg
% merged FindInPath() and __hbshell_FindInPath() into one
* contrib/hbxbp/xbpdialog.prg
+ Added: Main application windows object name as "PleseDoNotDelete"
* contrib/hbxbp/xbplistbox.prg
- Removed: traces from prev commit.
* contrib/hbqt/qtcore/hbqt_hbqevents.cpp
* contrib/hbqt/qtcore/hbqt_hbqslots.cpp
! Fixed: a couple of potential memory leaks.
* contrib/hbqt/qtcore/hbqt_bind.cpp
+ Implemented: Thread safe global list. Now objects are released
in a proper manner.
+ Added: a hack not to delete a QMainWindow if its name is
"PleaseDoNotDelete". It is scheduled to be removed
once it is resolved that why on certain circumstances, if
a QMainWindow is deleted, it results in GPF.
