* src/rtl/cdpapi.c
* use CP437 control codes in translations from unicode to ASCII
* src/rtl/gtwvt/gtwvt.c
* added support for key code typed using ALT + <ASCII_VALUE_FROM_KEYPAD>
* tests/gtkeys.prg
* added parenthesis to clarify expression order
* src/vm/dlmalloc.c
* use __builtin_clz() and __builtin_ctz() only with GCC 3.4 or newer.
Problem reported by David Arturo Macias Corona and OS2 GCC3.3.5.
* README.md
+ added link suggested by David to os2gcc454 from Paul Smedley
* src/rtl/gtstd/gtstd.c
+ accept keyboard input from redirected file on platforms without
dedicated code for keyboard support in GTSTD
* src/rtl/hbsocket.c
! use sock_errno() instead of errno in OS2 WATCOM builds.
It fixes few problems with socket code using errno value, i.e.
hb_socketConnect() with timeout.
* src/compiler/hbdbginf.c
! fixed module names generated for debugger line info when user compiles
files passing path with drive letters, i.e. hbmk2 c:\temp\test.prg
Thanks to Anton Ryzhov for the information about the problem.
* src/rtl/hbsocket.c
! fixed hb_socketConnect() with timeout in DOS Watt-32 builds.
Watt-32 sets socket error to EISCONN when connection is established.
* src/rtl/fstemp.c
! fixed casting for DOS C++ builds
* config/dos/watcom.mk
! moved CRTL library to the end of linked library list - some
other libraries may overload some C RTL functions or symbols,
f.e. error descriptions in WATT-32
* utils/hbmk2/hbmk2.prg
! added hbpmcom library to linked library list in DOS builds
! fixed WATT-32 support in DOS builds to work in cross compilation
environment
! link WATT-32 library before C RTL
! fixed simultaneous compilation with -jobs=N when N >= 2
The same hash array (hReplace) was shared between different threads
and each thread tried to modify it overloading their own parameters.
It was causing different side effects, i.e. all threads compiled
the same .c file instead of their own ones or HBMK2 was crashing
due to memory corruption.
* always join all started threads
* contrib/hbct/getsecrt.prg
! applied Clipper compatibility fix to GetSecret() by Pavel Tsarenko
* src/common/hbdate.c
* removed explicit _BSD_SOURCE definition - it was confusing our
code used to detect localtime_r() support and is not necessary
if we do not use extended fields of tm structure
* src/rtl/mtran.c
* updated to show strict Clipper compatible behavior - Clipper
ignores 3-rd cSoftCR parameter if 2-nd one cHardCR is not given
(strict compatibility is covered by HB_CLP_STRICT macro and
disabled by default - undocumented feature which may confuse
users)
% minor optimization
* contrib/gtqtc/gtqtc1.cpp
* accept META key as CTRL key in MacOSX builds.
This QT Meta key is marked as "control" on Apple keyboards so
this seems to be less confusing. Anyhow I'm not regular MacOSX
user so if someone thinks it's bad idea then please inform me.
"command" key is mapped by QT as CTRL key by default.
! disabled Qt::WA_InputMethodEnabled in default builds.
I disabled it because it was breaking support for national characters
in few European countries, f.e. Polish characters with ALT in MacOSX.
Qt::WA_InputMethodEnabled is used for some Asian languages but I have
no idea if it was working with GTQTC so far and if it's really
necessary and usable. I'm waiting for information from people who
lives in countries where it's used. Just let me know if you need
GTQTC compiled with this window attribute.
* contrib/gtqtc/gtqtc.hbp
* do not use system QT directories in android builds
* include/hbgtinfo.ch
* contrib/gtqtc/gtqtc1.cpp
+ added new font attributte HB_GTI_FONTA_NOSTRETCH
It disables font stretching in some GTs like QTC.
On some platforms (i.e. MacOSX) font stretching
does not work correctly (at least in my tests) so
if someone wants to use GTQTC on MACs then I suggest
to set:
hb_gtInfo( HB_GTI_FONTATTRIBUTE, ;
hb_bitOr( HB_GTI_FONTATTRIBUTE ), HB_GTI_FONTA_NOSTRETCH )
! do not use QPainter::RasterOp_SourceXorDestination to display
cursor and selection in MacOSX builds.
RasterOp operations are not supported in MacOSX.
* include/hbdate.h
* src/common/hbdate.c
+ added new C functions:
double hb_timeLocalToUTC( double dTimeStamp );
long hb_timeStampUTCOffset( int iYear, int iMonth, int iDay,
int iHour, int iMinutes, int iSeconds );
; warning: some C RTLs may not correctly detect Summer time for all past
dates in few countries.
* include/harbour.hbx
* src/rtl/dateshb.c
+ added new PRG function:
hb_TSToUTC( <tsLocal> ) -> <tsUTC>
+ added support for optional timestamp parameter in hb_UTCOffset()
function, current syntax is:
hb_UTCOffset( [ <tsLocal> ] ) -> <nSeconds>
* contrib/hbtip/hbtip.hbx
* contrib/hbtip/utils.c
+ added new PRG function:
tip_FileNameMimeType( <cFileName> ) -> <cMimeType>
It's a wrapper to old C function s_findFileMimeType()
TODO: replace few independent mimetype functions by single one.
! fixed tip_FileMimeType() to not call hb_fsClose() or any other FS
function with wrong handler.
! fixed tip_TimeStamp() using hb_timeStampUTCOffset() instead of
hb_timeUTCOffset() to correctly calculate UTC offset for dates
different then current one.
! decode time passed as numeric second() value in 2-nd parameter of
tip_TimeStamp() only when the 1-st one contains pure date value.
* src/debug/dbgentry.c
+ added protection against wrong parameters passed to __dbgDelBreak()
* src/debug/debugger.prg
+ implemented display history for command window.
This modification also fixes clearing command window when focus is
changed.
+ added support for new commands:
DELETE ALL BP
DELETE BP <nNumber>
LIST BP
* replicated Clipper compatible command line shortcuts
! send :RefreshAll instead of :RefreshCurrent to source code window
when break point changed
; TODO: add support for window resizing and repositioning. Commands
like LIST BP are in practice unusable in 3 line command window.
* src/rtl/console.c
% minor optimization
* src/debug/dbgtmenu.prg
! allow to chose current menu item by its hotkey
* src/debug/dbgtmitm.prg
% eliminated dummy hb_DispOutAtBox() for not marked options
* src/debug/debugger.prg
! keep RunAtStartup flag synced with menu
! fixed options decoding from init.cld file.
Thanks to Franček Prijatelj for locating the problem though fix
should be a little bit different.
* src/debug/tbrwtext.prg
! respect LineNumbers on/off setting
* src/rtl/console.c
! call hb_gtFlush() at the end of hb_DispOutAtBox() - it fixes
different side effects in code using hb_DispOutAtBox().
* contrib/hbct/ctwfunc.c
! extract box frame string passed to WBOX() using BOX CP
* src/vm/procaddr.c
! added hb_extIsNil() to hb_vmProcAddress() - it fixes PCODE DLLs
using [HB_]ISNIL() macro and linked with hbmaindllp library or
other wrapper using hb_vmProcAddress()
* src/vm/codebloc.c
- removed old comment which was never true in real HVM MT model
* ChangeLog.txt
! typo: tanks -> thanks ;)
* src/debug/debugger.prg
% Don't save debugger window positions if user did not change them as the
original debugger behaves.
+ Support arguments to Options Save and Options Restore debugger commands
as the original debugger does.
* src/debug/debugger.prg
! Maintain fixed window logical order as the original debugger does.
! Load debugger settings in two steps: before and after debugger UI activation.
! Save watchpoints to the settings file.
* src/debug/dbgentry.c
! added missing casting
* src/debug/debugger.prg
+ show in call stack references to code compiled without debug info
* src/rtl/itemseri.c
* ignore HB_HASH_RESORT flag during hash serialization
* src/debug/dbgtwin.prg
* src/debug/debugger.prg
! use explicitly INKEY mask in INKEY() calls to avoid interactions
with user code
* src/debug/debugger.prg
* src/debug/tbrwtext.prg
% removed code used for internal breakpoints handling - it was
redundant and all necessary information can be easy and fast
accessed directly from dbgentry.c. Current code is smaller,
faster and we do not have to worry about keeping break point
information synchronized between C level debugger internals,
HBDebugger object and HBBrwText object.
! changed source code window navigation key to work like in
Clipper debugger
! changed source lines displaying to work like in Clipper debugger -
with HOME/END scrolling code horizontally it's not such important
to always see line numbers
! redirect few missing K_CTRL_* keys to active window it fixes key
navigation in some of them
! redirect K_ENTER to command window if command is not empty just
like in Clipper debugger
* src/rtl/inkeyapi.c
! do not convert ASCII 127 to character in hb_inkeyKeyString()
and HB_KEYCHAR()
* src/debug/dbgentry.c
+ added new PRG functions:
__dbgIsBreak( <pDbg>, <cModule>, <nLine> ) -> <nPos>
__dbgGetBreakPoints( <pDbg> ) -> <aBreakPoints>
* src/debug/debugger.prg
- removed old and long time unused PRG function __dbgAltDEntry()
* src/rtl/hbproces.c
* changed returned status when execvp() fails, -1 should give
maximal exit code on given platform (i.e. 255)
* include/hbapidbg.h
* src/debug/dbgentry.c
* src/debug/debugger.prg
+ added new C function hb_dbgGetModuleName() - it returns module name
used to register module in HVM. In current version it only strips
path part though in the future it may use some more advanced code
which allow to debug modules with the same name but different paths
in single application and such extension will be local to core debug
code if upper level debugger uses hb_dbgGetModuleName()
+ added new PRG function __DBGGETMODULENAME() - it's wrapper to
hb_dbgGetModuleName()
+ added new PRG function __DBGMODULEMATCH()
! use __DBGMODULEMATCH() instead of hb_FileMatch() and strip_path()
! eliminated strip_path() from PRG code and use hb_dbgGetModuleName()
in C code in places where it's necessary
! replaced few strcmp() used to compare module names with
FILENAME_EQUAL()
; above modification may fix few potential problem with breakpoints
in debugger on some platforms if user uses different paths for
debugger and compilers.
! eliminated hack with memvar/field name copy in watch points,
it also fixes memory leak when fields are used in watch points.
% replaced few DO CASE with SWITCH
* contrib/gtqtc/gtqtc.h
* contrib/gtqtc/gtqtc1.cpp
* replaced harbour oriented code with dedicated QT solution
for XWindow synchronization and clipboard selection.
* doc/xhb-diff.txt
* updated section MULTI THREAD SUPPORT to make it more clear
for users and added information about SETKEY() and GTs.
* include/hbgtinfo.ch
+ added new hb_gtInfo() action: HB_GTI_RESIZESTEP
It enables/disables window resize progression.
* contrib/gtqtc/gtqtc.h
* contrib/gtqtc/gtqtc1.cpp
+ added optional code to enable XLib MT mode by XInitThreads()
+ added support for HB_GTI_RESIZESTEP and disabled window resize
progression in default settings - with some XWindow managers
it may cause bad side effects.
* do not center window in SETMODE() if console window dimensions
(rows and cols) are not changed.
* contrib/hbhpdf/3rd/libhpdf/hpdfimac.c
* modified isAligned() to not generate warnings on platforms were
'unsigned long' is not large enough to hold pointer (Win64).
It should pacify warnings reported by Maurizio.
* contrib/hbhpdf/3rd/libhpdf/libhpdf.dif
* rediffed with:
hbmk2 ../../../../bin/3rdpatch.hb -rediff
* utils/hbmk2/hbmk2.prg
! reverted 2nd hunk of 2013-06-03 16:45 UTC+0200. It
introduced unnecessary and confusing disparity between
command-line and .hbp/.hbm while they are supposed to
be working similarly. It also breaks syntax supported
in .hbc directives, and it also makes behavior contracting
with the way documented in hbmk2 help and hbmk2 examples.
All this, to resolve the problem of space containing option
values for one specific and rare subcase of reverted apostrophs.
To pass option/directive value containing a space, put
the whole option (or the option value) in double quotes
(as documented in help), this works equally well on the
command line and in .hbp/.hbm/.hbc files.
* ChangeLog.txt
! ending whitespaces
! explicitly filled git $Id
* include/hbpp.h
* src/pp/ppcore.c
* replaced compile time macro HB_PP_MULTILINE_STRINGS with
runtime #pragma setting, i.e.:
#paragma multilinestrings = on
* src/vm/hashes.c
% resize index during hash arrays sorting, i.e. when strict order
is disabled or user call HB_HSORT() function
! fixed typo in timestamp value comparison - in practice only date
part was significant
! fixed new code I added recently for resorting hash arrays with
strict order
* contrib/hbtip/client.prg
! Fixed: to test if a connection is available actually or not
before sending other info. This fixes a RTE when a SSL over
HTTP request was made and internet access was not available.
* contrib/hbtip/httpcli.prg
+ Added: METHOD Head() for HEAD verb of HTTP interface.
METHOD SetConnectionPersistent()
This does not send "Connection: close" header entry.
Useful in cases where many requests are required to
be submitted under one session. :close() may be
called to close the connection explicitly.
METHOD IsConnectionAlive()
It allows the application to test connection's state
and facilitates to take alternate action.
* src/vm/hashes.c
! remove index when natural order is disabled by
hb_hKeepOrder( hVal, .f. )
! do not update index when item is deleted and hash array
is marked to be resorted
% few minor optimizations
* contrib/sddoci/sddoci.hbc
* use ocilib as library name in Linux builds.
; I do not know on which platforms ocilib[wma] is used but it's
possible that this modification should be done also for other
platforms not only for Linux.
* src/rtl/gtwvt/gtwvt.c
! added workaround for AltGR and German keyboard
* src/rtl/inkeyapi.c
! added missing 'static' in function declaration
* src/rtl/hbgtcore.c
! added workarounds for HB_GTI_COMPAT_BUFFER switched to .F. by
default.
* tests/gtkeys.prg
* switched to UTF8 CP by default
+ extended keycode information in Harbour
* tests/mt/mttest10.prg
+ allow to chose GT passing it's name as command line argument
+ initialize GT in main thread if GT needs it
(currently QTC needs initialization in main thread)
* contrib/gtqtc/gtqtc1.cpp
* added workaround for missing MT synchronization in QT for XLIB
QT was designed to use screen widgets only in main thread so
it's only workaround which helps but I cannot guaranti
* contrib/hbmzip/mzip.c
* workaround for wrong st_mtime type in struct stat on android
* src/rtl/hbsocket.c
* do not check if IPPROTO_IP is defined, on some platforms like
android it's only enum type. Warning if it exploits some problems
on some seldom used platforms/compilers then please report it.
* src/vm/hvm.c
* minor
* package/mpkg_ver.sh
* do not use bash extensions
* package/harbour-wce.spec.in
* package/harbour-win.spec.in
* package/harbour.spec
* package/mpkg_rpm.sh
* package/mpkg_rpm_wce.sh
* package/mpkg_rpm_win.sh
* removed bash from dependencies list
* src/rtl/hbtoken.c
+ added support for passing token parser parameters as bit field number.
The goal is to replace last to parameters with single bit field which
allows to set many different parsing aspects instead of adding many
new parameters.
+ added support for reverted apostrophes quoting: `a b c`
; TODO: add constant values to control parser in token functions
* utils/hbmk2/hbmk2.prg
! respect reverted apostrophes quoting in options read from .hbp/.hbm
files.
! do not ignore tool parameters passed in -*flags= hbmk2 params if
they do not start with '-'. It resolves problems with shells which
need special quoting to pass parameters, i.e. with both modifications
it's possible to use in hbp files dynamic parameters encapsulated in
different forms, depending on used platform, shell and parameter type,
i.e. parameters are passed as set of separated arguments:
-cflag=`config-tool1`
or paramters are passed as single argument:
-cflag="`config-tool2`"
or parameters are passed in file which name shows config-tool3:
-cflag="@`config-tool3`"
In short words now things like:
-cflag=`pkg-config --cflags gtk+-3.0`
will work correctly with .hbp files
