* INSTALL
+ Added TROUBLESHOOTING/9.: how to submit logs to dev list.
Please read it.
* src/rtl/diskspac.c
* src/rtl/disksphb.c
! Fixed possible MT race condition in DISKSPACE()/HB_DISKSPACE()
where no value could be returned.
* src/rtl/gtwvt/gtwvt.c
* Changed to use WS_EX_COMPOSITED instead of WS_EX_LAYERED.
This works on XP and above (previous one also worked on W2K),
and I mean this as another round of live testing, as WS_EX_LAYERED
turned out to be causing strange flickers (maybe hardware
induced, I don't know) on some non-virtualized/non-remote
OS installations. Maybe this will work better. If not, even
this will be deleted. Then maybe software/DC double-buffering
will be the ultimate soltion.
% Dropped fiddling with SetLayeredWindowAttributes() call.
* Formatting.
* contrib/make.hbs
+ Added 'verbose' keyword. Equivalent to HB_BUILD_VERBOSE=yes.
+ Added 'rebuild' keyword. (Equivalent to 'clean all')
* contrib/3rd/sqlite3/sqlite3.hbp
! Added workaround for gcc/mingw bug in 4.5.0 and 4.5.1
revisions where it shows compile error when iso89 c dialect
is selected. Now in 4.5.x gnu89 is used instead to avoid it.
* contrib/hbplist
- Disabled gtwvg due to build breakage.
* include/harbour.hbx
* include/hbapi.h
* src/common/hbver.c
* src/rtl/version.c
+ hb_verHostCPU() to return CPU architecture the host OS
is running under. (long time TODO)
It may or may not give precise answer depending on
target OS. For win, it works as exptected.
+ HB_OSCPU() -> <cCPU> same for .prg level apps.
* contrib/hbmxml/3rd/minixml/mxml_str.c
! Ultimate isdigit() fix. [Tamas Tevesz]
* contrib/hbmxml/3rd/minixml/mxml.hbp
* contrib/hbmxml/3rd/minixml/mxml_str.c
! Fixed isdigit() warning which was result of missing
mxml header. [Tamas Tevesz]
+ contrib/hbcurl/hbcurlt1.ch
+ contrib/hbcurl/core_t1.c
+ Implemented new way of using HB_CURLOPT_HTTPPOST, now
all meaningful values are supported in any order and
combination. See:
http://curl.haxx.se/libcurl/c/curl_formadd.html
The patch is heavily based on patch/implementation submitted
by Benjamin Jacobs, many thanks for it. Final patch
includes some rework, simplification, compile and
other fixes, plus formatting with uncrustify. Please
make sure to test/review this version as I didn't do
any functional tests, just build ones.
Sample code:
curl_easy_setopt( curl, HB_CURLOPT_HTTPPOST, { ;
{ HB_CURLFORM_COPYNAME, "nparam1", ;
HB_CURLFORM_COPYCONTENTS, "value1" }, ;
{ HB_CURLFORM_COPYNAME, "nparam2", ;
HB_CURLFORM_COPYCONTENTS, "value" }, ;
{ HB_CURLFORM_COPYNAME, "fileparam", ;
HB_CURLFORM_FILE, filename } } )
WARNING: This implementation is INCOMPATIBLE with the
previous one.
TOFIX: Current implementation does allow to create GPFs because
it allows .prg code manually pass buffer sizes. This must
be fixed and buffer length should always be automatically
taken from Harbour item. This will require more sophisticated
code to build the low-level array (f.e. size calculation).
Benjamin, can you look at it?
NOTE: I'm uploading this unfinished code as new temp files,
because SVN doesn't support light branching, so this poor man's
solution is left to manage it.
* harbour/src/vm/procaddr.c
! removed dummy entry at the end of function table
* harbour/src/common/hbstr.c
* harbour/src/rdd/dbfcdx/dbfcdx1.c
* harbour/src/rdd/hsx/hsx.c
* harbour/contrib/xhb/fparse.c
* pacified warnings
* harbour/include/hbsetup.h
* disabled flatten function attribute in ANDROID builds,
it causes that Android_NDK-GCC GPFs compiling src/vm/procaddr.c
* harbour/include/hbdate.h
* enabled localtime_r() in Android builds
* harbour/src/common/hbgete.c
* enabled setenv() in Android builds
* harbour/src/common/hbtrace.c
! fixed ANDROID_LOG_* flags
; QUESTION: why Android log is used instead of syslog?
It causes that harbour has to be linked with
log library (-llog).
* harbour/src/common/hbprintf.c
* disabled modfl() in Android builds
* harbour/src/rtl/filesys.c
* added hack for missing functions in android libc library
* harbour/src/rtl/diskspac.c
* harbour/src/rtl/disksphb.c
* use statfs instead of statvfs in Android builds
* harbour/contrib/hbct/dattime3.c
* disabled stime() in Android builds
* harbour/contrib/hbmzip/3rd/minizip/ioapi.h
* do not use 64bit FILE functions in Android builds
* harbour/contrib/xhb/hbsyslog.c
* disabled using Android log in Android builds - use standard syslog
instead. This code needs deeper modification to switch to Android log.
; I have just recompiled whole Harbour code for android-3, android-5
and android-9 at ARM using standard Harbour Linux build files.
Android does not have pthread and rt libraries (they are merged
with android c library) so to avoid link errors I simply created
soft links from libpthread.a and librt.a to libc.a.
With this hack could use unmodified hbmk2 and linux .mk files.
BTW exporting HB_CPU="arm" was not enough to switch to cross build
in my x86@64 Linux machine so I had to export HB_HOST_BIN and set
HB_BUILD_3RDEXT="no"
* harbour/include/hbapifs.h
* harbour/src/rtl/filesys.c
+ added new C function:
int hb_fsIsPipeOrSock( HB_FHANDLE hPipeHandle );
* harbour/src/rtl/filesys.c
+ added protection against slow function interrupting by signals
in *nixes
* harbour/include/hbapifs.h
* harbour/src/rtl/filesys.c
+ added new C functions:
HB_BOOL hb_fsPipeCreate( HB_FHANDLE hPipe[ 2 ] );
HB_SIZE hb_fsPipeIsData( HB_FHANDLE hPipeHandle, HB_SIZE nBufferSize,
HB_MAXINT nTimeOut );
HB_SIZE hb_fsPipeRead( HB_FHANDLE hPipeHandle,
void * buffer, HB_SIZE nSize,
HB_MAXINT nTimeOut );
* harbour/src/rtl/philes.c
+ added new PRG function:
HB_PREAD( <nPipeHandle>, <@cBuffer>, [<nBytes>], [<nTimeOut>] )
-> <nBytesRead>
It can be used for unblocking read from pipes, i.e. using handles
returned by hb_processOpen() function.
<nPipeHandle> is pipe read handle
<cBuffer> is string variable passed by reference with read buffer
<nBytes> is maximum number of bytes which can be read
<nTimeOut> is timeout in milliseconds to wait for data to read,
this function returns when at least one byte can be read
<nBytesRead> is number of bytes read from the pipe,
on error this function return -1.
* harbour/src/rtl/hbproces.c
+ add new implementation of hb_fsProcessValue() and hb_fsProcessClose()
for OS2 which uses OS2 API
* use hb_fsPipeCreate()
* harbour/src/rtl/hbjson.c
* minor code reorganization to cleanup warning
* include/hbapi.h
* include/harbour.hbx
* src/common/hbver.c
* src/rtl/version.c
+ added hb_verHostBitWidth( void ) -> int which retrieves bit width of host OS
+ added HB_OSBITWIDTH() -> <nBits>
; It will currently work on win platform, falling back to target arch bits
for others. Pls extend it for *nixes. For MS-DOS, 32 will be returned with
both currently supported compilers.
* harbour/src/vm/thread.c
* renamed hb_criticalCode() to hb_mutexEval()
* harbour/src/rtl/cdpapi.c
* added HB_CODEPAGE_ANNOUNCE( UTF8 ) so user can use
REQUEST HB_CODEPAGE_UTF8
in PRG code.
* harbour/tests/big5_gen.prg
; added note about characters mapped to U+FFFD REPLACEMENT CHARACTER
in BIG5.TXT and iconv mapping for the same characters.
* harbour/src/rtl/cdpapi.c
! fixed U+0000 conversion - it should not be converted.
* harbour/src/rtl/hbjson.c
! fixed decoding unicode value from strings with \uHHHH
! fixed to use hb_cdpU16ToStr() instead of hb_cdpGetChar()
which cannot be used with CPs using multibyte encoding.
* harbour/ChangeLog
! fixed typos in my previos ChangeLog entry
(I should try to read such things at least once before I'll commit)
; TOFIX: contrib/hbexpat/internal.c: uses hb_cdpGetU16() which cannot
be used with CPs using multibyte encoding.
* harbour/include/hbapicdp.h
* harbour/include/hbcdpreg.h
* harbour/src/rtl/cdpapi.c
* harbour/src/rtl/cdpapihb.c
+ added support for user defined character encoding.
Now it's possible to easy create and register in HVM CPs using any
encoding so I expected that users interested in adding support for
some exotic character encoding will create such mapping instead
of asking for it.
% moved support for codepages using multibyte characters mapped to
single unicode values to user defined encoding - it nicely simplify
the code and eliminate unnecessary overhead in other CPs.
* allow to chose UTF8 as HVM CP
* harbour/src/rtl/idle.c
! fixed idle mode flag resetting
* harbour/src/rdd/hbsix/sxutil.c
! fixed SX_SLIMFAST() results when nested quoting with (") and (') is
used
+ harbour/tests/big5_gen.prg
+ added code to generate C source with conversion tables between
BIG5 and UCS16 using data defined by Unicode, Inc. in BIG5.TXT
+ harbour/src/codepage/cp_utf8.c
+ added alternative UTF8 Harbour CP (UTF8ASC) as an example Harbour
user defined codapged using multibyte character encoding
* harbour/src/codepage/Makefile
+ harbour/src/codepage/cp_u16le.c
+ added Harbour codepage using UTF16 little endian encoding
* harbour/src/codepage/Makefile
+ harbour/src/codepage/big5.c
+ harbour/src/codepage/cp_big5.c
+ added BIG5 Harbour CP. It can be used with programs using Harbour
STR API with automatic translations.
; This CP needs really big translation tables. I added code which
makes some very simple compression which reduced raw size from
176100 bytes to 77354 but it's still large 77KB so maybe we should
think about moving this CP to other Harbour codpage library which
is not part of harbour shared library harbour*{.dll|.so|.dyn|...}
Alternatively I can try to reduce static size to about 30KB and
then build necessary tables dynamically at runtime when they are
used first time though in such case I will need additional 177KB
of dynamic memory instead of 77KB of static memory used by current
code.
* src/rtl/fstemp.c
! hb_fsTempDir(): due to misplaced else branche it returned
'./' on *nix systems when temp dir ended with pathsep,
for the same reason it return empty string instead of './'
if temp dir could not be determined.
; Please check me.
* src/rtl/gtxwc/gtxwc.c
% undone small optimization in 2011-04-04 10:48 UTC+0200 Przemyslaw Czerpak
because it caused random app stalls where the app would only
continue if some input events were manually made. Cursor
blinking also only happened on input events.
I've seen this issue on Ubuntu 10.10 and latest OS X.
After this patch, it works OK again.
Przemek, please review this patch, most probably there is
a solution which keeps the optimization and avoids this problem.
* harbour/src/rtl/gtxwc/gtxwc.c
% small optimization
! added hack for multiple window resizing when user executes
series of setmode() function. I'll look for cleaner solution
in some spare time.
* src/rtl/gtwin/gtwin.c
% Replaced kbd modifier readout with a more efficient
version. It contains hack required to make core
low-level function to work correctly with win consoles.
Patch co-developed with Przemek.
* contrib/hbtip/sendmail.prg
! Implemented fix to enclose attachment filenames in
dbl quotes as suggested by George Frehner.
Pls test.
* contrib/hbtip/mail.prg
* Formatting.
* src/rtl/gtwin/gtwin.c
- Fully disabled palette setting feature due to bad side effects
reported by Mindaugas.
Maybe it could be reenabled in some cases, pls refine it if you
need this feature. Pls also retest, I have a hard time replicating
this side-effect on my system.
* INSTALL
* Updated. mingw tdm 4.5.2 works fine again.
* examples/hbdoc2/gentpl.prg
! Fixed typo reported on users' list by Emmer.
* harbour/src/rtl/hbznet.c
% use Z_PARTIAL_FLUSH instead of Z_SYNC_FLUSH in hb_znetFlush()
It should reduce a little bit the size of compressed data stream.
* harbour/include/hbapi.h
* harbour/src/rtl/hbrandom.c
added new C functions which fills memory block with random values:
void hb_random_block( void * data, HB_SIZE len )
this function should be used with "good enough" random generator
so in the future it may be changed to use internally sth different
then rand().
* harbour/src/vm/thread.c
* pacified minor warning
* harbour/src/rtl/hbproces.c
! fixed one HB_TRUE wrongly used instead of TRUE
* minor simplification
* harbour/include/hbgtinfo.ch
* harbour/src/rtl/hbgtcore.c
+ added HB_GTI_VERSION - it returns version information just like
hb_gtVersion() - usually the same but can be different in some
multilevel GTs, i.e. when different GT drivers are used locally
and remotely.
* harbour/include/inkey.ch
* harbour/src/rtl/hbgtcore.c
+ added HB_K_CONNECT and HB_K_DISCONNECT inkey codes.
They can be used to signal that remote terminal is connected or
diconnected.
* harbour/tests/gtkeys.prg
+ added HB_K_* keycodes
* harbour/contrib/hbnetio/netiosrv.c
* removed unnecessary casting
* harbour/src/rtl/filesys.c
! added protection for potentially unfinished due to signal
interrupt close() operation in *nix systems in hb_fsClose()
function
* use hb_fsClose() instead of close() to be sure that the operation
is completed in *nix systems
* return NULL for DosToWinHandle( FS_ERROR )
; TODO: add protection against signal interrupting to all hb_fs*()
functions which may need it.
* harbour/src/rtl/hbproces.c
* use hb_fsClose() instead of close() to be sure that the operation
is completed in *nix systems
! Do not inherit ends of pipe handles in MS-Windows builds of
hb_fsProcessOpen()
! Added missing CloseHandle( hProcess ) in MS-Windows builds of
hb_fsProcessRun()
; QUESTION: Does hb_fsProcessRun() work correctly in MS-Windows builds?
I've just noticed that I used WaitForMultipleObjects()
WIN API function with unnamed pipe handles and I cannot
find any information in MSDN that this is legal operation.
* harbour/include/hbgtinfo.ch
+ added HB_GTI_ONLINE - Is terminal connected?
* harbour/src/rtl/hbproces.c
! fixed MS-Windows version of hb_fsProcessRun() to unlock HVM stack
when thread is waiting for child process
* harbour/src/rtl/hbrandom.c
! fixed double number calculation for random values by removing
some wrong trick with epsilon value. rand() function returns
numbers in range [0,RAND_MAX] (inclusive) so can give RAND_MAX+1
different results and this is the correct divisor to map rand()
results to real range [0,1) ( 0 <= n < 1 )
* harbour/src/lang/msgbgiso.c
! fixed typo in lang module name: "BGWIN" -> "BGISO"
; TODO: sync codepage and land module names to use the same ID.
Now it's necessary to look at Harbour source code to find
corresponding modules.
* harbour/src/rtl/gtwvt/gtwvt.c
! fixed GPF when fullscreen was enabled before console window is
created.
Now it's possible to enable full screen mode before application
starts without flickering window, i.e.:
#include "hbgtinfo.ch"
REQUEST HB_GT_WVT_DEFAULT
PROC main()
hb_gtInfo( HB_GTI_ISFULLSCREEN, .T. )
? hb_datetime()
? version()
? hb_compiler()
? hb_gtVersion(), hb_gtVersion(1)
WAIT
RETURN
* src/rtl/hbzlibgz.c
! In block declaration in recent patch.
* utils/hbmk2/hbmk2.prg
! Regression in recent xhb patch.
* contrib/hbwin/axcore.c
! Format warning in trace.
* src/rtl/gtwvt/gtwvt.c
* Indenting.
* INSTALL
+ Added new ghostscript link. [Tamas Tevesz]
(left the old one as it has some information the new one doesn't)
* contrib/hbqt/hbmk2_qt.hbs
* contrib/hbqt/qtgui/qth/*.qth
* contrib/hbqt/qtcore/hbqt_pointer.cpp
* contrib/hbqt/qtcore/hbqt.h
* contrib/hbqt/qtcore/qth/*.qth
* contrib/hbqt/qtnetwork/qth/*.qth
* contrib/hbqt/qscintilla/qth/*.qth
* contrib/hbqt/qtdesigner/qth/*.qth
+ Patch to move wrapper code to C++. [Francesco Perillo]
* My mods:
- Restored the way registerResource() is implemented.
(TODO: needs generic solution)
- Formatting, some cleanups.
- Minor corrections in generated code format (more TBD)
- Deleted some unused code
- Refixed strcpy() usage.
- Added C++ puller for __QT* symbol.
- Added manual version of HBQEventsFromPointer() which is still needed
by hbqtcore. (TOFIX)
Please test.
; TOFIX: Remove remaining generated .prg, holding this code:
FUNCTION QtClass( ... )
RETURN HB_QtClass():new( ... )
Any takers?
* harbour/src/rtl/minmax.c
! fixed MAX() for mixed date and timestamp values
* harbour/src/rtl/strtran.c
% minor modification
* harbour/src/rtl/cdpapihb.c
* updated for signed HB_SIZE type
* harbour/src/rtl/valtype.c
* modified to work with any HB_IT_STRING flags we may add in the future
* harbour/src/rdd/dbcmd.c
* eliminated explicit hb_parinfo() usage
* harbour/src/compiler/harbour.y
* added casting which seems to be necessary for some newer bison versions
* harbour/doc/xhb-diff.txt
* typo
* harbour/config/win/bcc.mk
* harbour/utils/hbmk2/hbmk2.prg
* disabled W8060 warnings in BCC builds when warning level is set to low
* harbour/src/rtl/gtwin/gtwin.c
! added missing void in hb_gt_win_IsFullScreen() declaration
* harbour/src/rtl/hbjson.c
! fixed typo which caused that \uNNNN in strings were ignored
instead of being decoded
; QUESTION: why we are not using unicode strings in encoding by default?
* include/hbgtinfo.ch
* src/rtl/gtwin/gtwin.c
* src/rtl/gtwvt/gtwvt.c
+ HB_GTI_ISFULLSCREEN checking and setting inside GTWIN done.
! Minor fix for GTWVT, code patched to work on Windows 95 (tested).
; [Aleksander Czajczynski]
* config/beos/gcc.mk
* config/bsd/clang.mk
* config/bsd/pcc.mk
* config/linux/clang.mk
+ add soname specs to some gnu lds [Tamas Tevesz]
* harbour/utils/hbmk2/hbmk2.prg
* harbour/config/dos/watcom.mk
* harbour/config/win/watcom.mk
* harbour/config/linux/watcom.mk
* harbour/config/os2/watcom.mk
! fixed to not use -wcd124 and -wcd136 in OpenWatcom C++ builds
* harbour/config/win/xcc.mk
! added -noexpobj as workaround for problems with creating
shared library
* removed now unnecessary $(RM) harbour*.dll
* harbour/src/rtl/base64d.c
* pacified warning
% removed unnecessary condition with RTE - decoded string has to
be shorter then the source one taken from HVM string item so for
sure it cannot exceed maximum string item size
* harbour/src/rtl/base64c.c
* generate RTE if encode string size is too big
* eliminated unnecessary INT_MAX string limit
* include/hbgtinfo.ch
* src/rtl/hbgtcore.c
* src/rtl/gtdos/gtdos.c
* src/rtl/gtwin/gtwin.c
* src/rtl/gtxwc/gtxwc.c
* src/rtl/gtcrs/gtcrs.c
* src/rtl/gtstd/gtstd.c
* src/rtl/gttrm/gttrm.c
* src/rtl/gtos2/gtos2.c
* src/rtl/gtsln/gtsln.c
* src/rtl/gtpca/gtpca.c
* src/rtl/gtwvt/gtwvt.c
* contrib/hbct/ctwin.c
* contrib/gtwvg/gtwvg.c
* contrib/gtwvg/wvggui.c
* contrib/gtalleg/gtalleg.c
* contrib/hbrun/hbrun.prg
* tests/wvt_fs.prg
* examples/gtwvw/gtwvw.c
* Renamed HB_GTI_FULLSCREEN to HB_GTI_ISSCREENPOS
HB_GTI_FULLSCREEN is now deprecated and will be deleted in future
version of Harbour, use the new name instead.
* Renamed HB_GTI_FULLSCREEN to HB_GTI_ISFULLSCREEN
in context of GT window being fullscreen on physical display
recently added in 2011-02-26 20:07 UTC+0100 Viktor Szakats
; If someone has further name suggestions, pls tell.
* src/rtl/hbgtcore.c
* Setting defaults for HB_GTI_ISFULLSCREEN and HB_GTI_ALTENTER.
* include/hbgtinfo.ch
* src/rtl/gtwvt/gtwvt.h
* src/rtl/gtwvt/gtwvt.c
+ tests/wvt_fs.prg
+ Added ability to switch GTWVT screen into fullscreen. See example
for usage. Patch by Aleksander Czajczynski. Many thanks!
Pls test.
* contrib/hbqt/qtcore/hbqt_pointer.cpp
! Fixed bug introed in 2011-02-24 18:03 UTC-0800 Pritpal Bedi
where char array declaration used runtime size value.
Replaced with hb_strdup(). In previous commit I've also fixed
an strcpy() usage. strcpy() is strictly forbidden in Harbour
codebase. Hopefully it fixes MSVC build failure.
* src/rtl/fscopy.c
! Use hb_fsReadLarge()/hb_fsWriteLarge().
+ Added TODO to use hb_fileExtOpen().
! Deleted FXO_DEFAULTS after carefully checking there is no
more "extra" feature of hb_fsExtOpen() enabled in this specific
scenario.
* config/postinst.hbs
* config/global.mk
* config/bin.mk
* config/darwin/gcc.mk
* config/darwin/icc.mk
* config/darwin/clang.mk
* config/linux/gcc.mk
* config/dyn.mk
* config/os2/gcc.mk
* Trying to cleanup the harbour dynlib name situation.
Here's the plan (which is similar to what's used in contrib area):
win, wce: harbour-21[-subtype][.dll/.lib]
dos, os2: harbour[.dll|.???]
darwin:
libharbour.2.1.0.dylib
libharbour.2.1.dylib -> (symlink) [compatibility level]
libharbour.dylib -> (symlink)
*nix:
# libharbour.s?.2.1.0
# libharbour.s?.2.1 -> (symlink) [soname]
# libharbour.s? -> (symlink)
; It's possible it's broken now. Pls test linux/gcc and darwin.
'install' was not tested.
; TODO: Clean variable usage, there is some redundancy, plus
some places where current solution is not generic, f.e.
using HB_VER_*, HB_DYNLIB_BASE vs. HB_DYNLIB_NAME, etc.
* harbour/src/rtl/fscopy.c
* Reverted 2011-02-22 12:27 UTC+0200 Mindaugas Kavaliauskas
which made behavior inconsistent with rest of similar
functions like FERASE(), FRENAME(), which also don't throw
RTE if bad parameter is passed, but return FERROR()
and failure instead.
Also restored _SET_DEFAULT handling to not create a special case
compared to __COPYFILE() behavior, ia. some features like
FXO_SHARELOCK are still enabled while FXO_DEFAULT is not).
Pls rewrite it using hb_fsOpen()/hb_fsCreate() if that behavior
bothers you.
