* contrib/hbmzip/hbmzip.c
+ Added .prg level functions:
HB_ZIPSTOREFILEHANDLE():
To add a new file to a .zip via an already open handle (typically a temp file).
Obviously, file date/time/attrib will not be set.
HB_UNZIPEXTRACTCURRENTFILETOHANDLE():
To extract a file to an already open file handle.
; NOTE: I'm not sure if these new functions are useful
enough to be added to official SVN. Except for Windows,
there doesn't seem to be a trend that OSes would
support native temp files, so that we can use this
to avoid race conditions. Opinions are welcome.
* source/rtl/net.c
* contrib/hbziparc/hbziparc.prg
* contrib/hbct/files.c
* Minor formatting.
* source/rtl/gtxwc/gtxwc.c
! Fix to recent fix. An '!' was missing, Przemek pls check me.
* source/rtl/philes.c
+ HB_PROGNAME() now returns absolute path in *NIX systems, too.
(please test and refine)
* source/lang/msgca.c
! Fixes from xhb.
* contrib/hbziparc/hbziparc.prg
* Minor correction ( = -> := )
* source/rtl/gtwin/gtwin.c
+ Added HB_GTI_CODEPAGE support.
As per MS docs, this works only when non-raster font is used
in the console window. Notice that the accepted CP values are
different from the GTWVT ones (GTWIN being the more "logical"
one, since it's a newer API). Currently, Harbour doesn't try
to hide these interface details.
; TODO: UNICODE and BOX char support for GTWIN mode.
* source/rtl/tbrowse.prg
* Using HB_DISPOUTATBOX() to draw column/header/footer separators.
; NOTE: This may cause problems if someone wants to use native
CP (accented or other special CP specific) chars to draw
these screen elements. Maybe an option should be added,
or some sort of markings in the separator strings to
control that.
Pls REVIEW.
* include/hbextern.ch
* source/rtl/console.c
+ Added HB_DISPOUTATBOX() which is similar to HB_DISPOUTAT(),
but marks the text drawn as HB_GT_ATTR_BOX, so these chars
can be properly displayed as drawing chars, regardless of
the selected codepage. Another difference is that this
function only supports strings to be printed. Other types
don't print anything.
; NOTE: Maybe this function could use a better name, pls
review and decide.
* source/rtl/gtwvt/gtwvt.c
* source/rtl/gtwvt/gtwvt.h
! Fixed not accepting zero as a HB_GTI_CODEPAGE value.
! Fixed HB_GTI_CODEPAGE to change the codepage.
+ Added box char support in Unicode mode.
For chars marked as HB_GT_ATTR_BOX, chars will be
mapped to Unicode according to CP437.
+ Added box char support in non-Unicode mode.
For chars marked as HB_GT_ATTR_BOX, OEM_CHARSET will
always be used.
; NOTE: This way it's possible to use ISO/WIN codepages with
GTWVT, while still being able to use various drawing
chars (lines, blocks, arrows).
; NOTE: The available set of drawing chars depends on the
OS OEM_CHARSET mapping _for non-Unicode Harbour_.
For Unicode Harbour, full CP437 codepage is
always available.
; NOTE: This solution doesn't go as far as (GTWXC) to
graphically draw / emulate these chars.
; TODO: Add HB_GTI_BOXCP support to change hard-wired "EN" CP.
* include/hbapigt.h
* include/hbgtcore.h
! Moved HB_GT_ATTR_* macros to public headers, because they
are needed for public API functions.
* contrib/hbwin/win_reg.prg
+ w32_regRead(): Added second parameter to specify the default
value returned in case the reg entry doesn't exist.
If not specified, NIL will be returned, just like before.
+ GetRegistry(): Added 4th parameter with same purpose as above.
* source/rtl/hbregex.c
* Minor opt.
* contrib/hbwin/win_misc.c
* Added command line option parameter to WIN_RUNDETACHED().
* include/hbapi.h
+ HB_ERRCODE (to replace ERRCODE)
+ HB_SUCCESS (to replace SUCCESS)
+ HB_FAILURE (to replace FAILURE)
+ contrib/hbwin/win_misc.c
* contrib/hbwin/common.mak
* contrib/hbwin/Makefile
+ WIN_SHELLEXECUTE()
WIN_RUNDETACHED()
WIN_LOADRESOURCE()
Added some new Windows API wrappers.
; TODO: Add hb_osEncode() to them.
* source/rtl/diskspac.c
* source/rtl/disksphb.c
! Fixed DISKSPACE() and HB_DISKSPACE() for Darwin.
Previously they had returned erroneous values.
; TOFIX: Probably HB_DISKSPACE() should be implemented for Watcom and CEGCC,
just like DISKSPACE() is.
* source/rtl/filesys.c
! hb_fsGetAttr() to set the returned attribute to zero
in case of error.
* include/hbextern.ch
* source/rtl/philes.c
+ HB_FSETATTR( <cFileName>, @<nAttr> ) -> <lSuccess>
+ HB_FGETATTR( <cFileName>, <nAttr> ) -> <lSuccess>
+ HB_FSETDATETIME( <cFileName>, [<dDate>], [<cTime HH:MM:SS>] ) -> <lSuccess>
* source/rtl/direct.c
! Fixed DIRECTORY() when called with "V" (label) parameter.
Tested under Windows. Please test/correct for other platforms.
On Windows, the dirspec should not contain a filemask in order
to work.
* contrib/hbtip/base64x.c
% HB_BASE64(): Optimized, cleaned.
! HB_BASE64(): Removed second parameter allowing to pass the length of
the string. This could cause GPF if passed incorrectly. Now length
is simply determined using hb_parclen().
; TOFIX: Input string size limit checking.
* include/hbdefs.h
+ HB_SIZEOFARRAY() macro, which does: ( sizeof( var ) / sizeof( *var ) )
To be really precise, f.e. all hb_strncpy() calls should use
'hb_strncpy( d, s, HB_SIZEOFARRAY( d ) );' instead of:
'hb_strncpy( d, s, sizeof( d ) );' (given that 'd' is allocated at compile time).
So that the code adapts to changing character sizes (like UTF-16/32).
It's useful in some other places, too.
* contrib/hbct/dattime2.c
* 0 -> FALSE for BOOL types.
* include/hbextern.ch
* source/rtl/strmatch.c
+ HB_WILDMATCHI(). Case-instenstive, exact match. First
two parameters and return value are the same as for
HB_WILDMATCH().
* include/hbextern.ch
* source/rtl/cdpapi.c
+ HB_CDPSELECT() which does the same as HB_SETCODEPAGE(), but
aligns well with function naming rules and namespace.
HB_SETCODEPAGE() still works, but usage is not recommended.
+ Added HB_CDPUNIID( <cHarbourCP> ) -> <cCPName>
This will return the std CP ID of a Harbour CP, or empty
if the Harbour CP isn't linked.
* include/hbapi.h
* source/rtl/is.c
! Fixed compile error when HB_CDP_SUPPORT_OFF is defined.
+ Added following functions:
hb_charIsDigit()
hb_charIsAlpha()
hb_charIsLower()
hb_charIsUpper()
* include/hbapi.h
- hb_strUpperCopy() removed. Implementation was missing.
* source/rtl/strcase.c
* Minor formatting.
* source/rtl/filesys.c
! hb_fsCurDirBuff() fixed potential buffer overrun by one byte.
* source/rtl/hbrunfun.c
* source/rtl/run.c
! Fixed missing header hbapiitm.h (after previous local change).
* include/hbapi.h
* include/hbset.h
* include/hbapifs.h
* include/set.ch
* source/rtl/hbffind.c
* source/rtl/gete.c
* source/rtl/filesys.c
* source/rtl/hbrunfun.c
* source/rtl/run.c
* source/rtl/philes.c
* source/vm/set.c
* source/vm/cmdarg.c
+ Added file system CP translation.
Set( _SET_FNAMECP[, <cCPID> ] ) -> <cOldCPID>
All operations passing filenames to/from the
OS will convert them to/from the host CP from/to
the FS CP specified using _SET_FNAMECP.
* hb_fsNameConv() extended to deal with CP
conversion, if requested by the app.
+ hb_fsNameConvFrom() (exported) added to deal
with CP conversions for filenames received from OS.
+ hb_fsNameConvTo() (exported) added to deal
with CP conversions for filenames/string sent to the OS.
+ Added hb_setGetFNAMECP() (exported) function.
+ Added hb_setGetFileCPTransTo(), hb_setGetFileCPTransFrom()
(non-exported) functions.
+ hb_fsCurDirBuff() and hb_fsFindNextLow() extended
to use hb_fsNameConvFrom().
+ Added hb_fsBaseDirBuff() to return the base
directory on the C level (based on argv[0]).
+ Added HB_PROGNAME() to return the executable
program name (based on argv[0]).
+ Added HB_DIRBASE() to return the executable
base directory (based on argv[0]).
* Renames done regarding OS/FS CP conversion:
hb_fsNameConvFrom() -> hb_osDecode()
hb_fsNameConvTo() -> hb_osEncode()
hb_setGetFileCPTransTo() -> -
hb_setGetFileCPTransFrom() -> -
hb_setGetFNAMECP() -> hb_setGetOSCODEPAGE()
_SET_FNAMECP -> _SET_OSCODEPAGE
* hb_osDecode()/hb_osEncode() extended to have a second
parameter BOOL * fFree.
Changed all calls to handle the case when fFree is set
to TRUE by hb_osDecode()/hb_osEncode().
+ HB_GETENV(): Added 3rd logical parameter to control wether
to convert the returned value from OS CP to Harbour CP.
The default is TRUE to be in sync with GETE[NV]() and the
rest of core.
* hb_fsNameConv() is now doing OS CP conversion using std APIs,
rather than duplicating such logic. I've traded some speed
for modularity.
* hb_osDecode/hb_osEncode() function declarations moved to hbapi.h
* hb_osDecode/hb_osEncode() function definitions moved to set.c
; NOTE: None of these were extensively tested, and I tend
to make mistakes when dealing with pointers.
It compiles cleanly with BCC, and some basic tests
worked, but bFree = TRUE codepaths are not yet
working so this needs more review.
* source/rtl/teditor.prg
* source/rtl/tpersist.prg
* utils/hbmake/hbmake.prg
* MemoRead() -> hb_MemoRead()
* contrib/hbct/files.c
; TOFIX: Many functions here are possibly not thread safe.
; TOFIX: Calling hb_fsFindClose() at app exit (and thus keeping
the FF handles open for the whole app lifetime) doesn't
play well with server side apps (as the app may run for
months or even longer), and even worse, they
will lock the passed directory so that they cannot be
removed (maybe even renamed), until the app exists.
At least on Windows.
* source/rtl/dirdrive.c
! Fixed to use F_ERROR instead of -1.
* include/hbextern.ch
! Added missing HB_GCSTEP().
* source/vm/extrap.c
+ Added CPU dump code for Windows x64.
+ Enabled module listing for Windows 64-bit platforms.
% Minor cleanup in Windows x86 CPU dump code.
+ Added TODO for Windows IA64 CPU dump.
+ Added TODO for Windows x64 stack walk, with pointers.
+ Added TOFIX regarding displaying module names in Windows x64 mode.
* source/rtl/fstemp.c
! Minor formatting.
* source/rtl/gete.c
* source/rtl/run.c
* source/rtl/hbrunfun.c
* source/vm/cmdarg.c
+ Added CP conversion to:
GETE[NV](), HB_GETENV(), __RUN(), HB_RUN(),
HB_ARGSTRING(), HB_CMDLINE().
; TODO: dbCreateTemp( <cAlias>, <aStruct>, <cRDD>, <cDelimArg>, <nConnection> ) -> <lSuccess>
Generates and opens a new dbf with a temporary filename
in exclusive mode, deletes it automatically when closed.
All indexes created for this table, should ideally have
the temporary flag on, so those get deleted too, when closed.
; TOFIX: hb_regex*() functions will not honor Harbour CP setting,
when case insensitivity is selected.
; TOFIX: hbct / SetFDaTi() to move any local logic into Harbour core
functions and use a simple Harbour API call only.
I'm not sure what is the reason with keeping the logic
duplicated here; if this isn't by intent, this function
could be much simplified.
; TOFIX: Avoid using C level toupper()/tolower()/islower()/isupper()
; TOFIX: hb_regexMatch()'s 3rd parameter has a double meaning.
It controls case-sensitivity (in sync with the documentation),
but it also controls if hb_regexMatch() behaves like
hb_regexLike() (.T.), or hb_regexHas() (.F.). This means
that hb_regexMatch(,, .T.) == hb_regexLike(,, .T.) (case-sensitive) and
hb_regexMatch(,, .F.) == hb_regexHas(,, .F.) (case-insensitive)
Maybe this was the intent of the original developers, but
to me it looks a rather strange behaviour which was added by
accidentally forgetting about the 3rd parameter already being
utilized. This either needs to be fixed, or properly documented.
; NOTE: Noteworthy links:
http://docs.python.org/3.0/whatsnew/3.0.html
- Unicode/text section
- Removed <> (use != instead)
http://sphinx.pocoo.org/
- Documentation format
http://fredeaker.blogspot.com/2007/01/character-encoding-detection.html
- Character encoding detection
* harbour/source/rtl/gtxwc/gtxwc.c
! added protection against inaccessible character cell value
what could cause accessing uninitialized memory and GPF
* harbour/include/hbrdddbf.h
! fixed HB_DIRTYREAD() macro not updated when RDDI_* settings were
converted from global to thread local
* harbour/source/rdd/dbf1.c
* harbour/source/rdd/dbfntx/dbfntx1.c
* harbour/source/rdd/dbfnsx/dbfnsx1.c
* harbour/source/rdd/dbfcdx/dbfcdx1.c
* harbour/source/rdd/dbffpt/dbffpt1.c
% replaced <RDD>NODE_DATA() and <RDD>AREA_DATA() macros defined
in each RDD by common for all DBF bases RDDs DBFNODE_DATA() and
DBFAREA_DATA() macros defined in hbrdddbf.h to reduce possible
mistakes in code updating in the future
* harbour/contrib/rddads/adsx.c
* improved code to solve RDD initialisation order problem
* harbour/contrib/rddsql/sqlbase.c
! fixed minor typo
* harbour/source/pp/hbpp.c
* harbour/source/compiler/hbusage.c
* updated copyright year to 2009
* harbour/contrib/gtwvg/gtwvg.c
! Synchronized with GTWVT as per change below ( Przemek ):
+ Added 'layered' window support. It's enabled on W2K
and above (and only when not in Terminal Services
environment, because it makes it perform worse). This
way the screen is also properly displayed regardless
of what the Harbour application does. It also fixes
numerous paint artifacts, when GTWVT window was in
the background and another window is moved in front
of it.
* harbour/common.mak
* harbour/source/rtl/Makefile
* harbour/include/hbapifs.h
* harbour/include/hbextern.ch
+ harbour/source/rtl/hbproces.c
+ harbour/source/rtl/hbprocfn.c
+ added C functions hb_fsOpenProcess(), hb_fsProcessValue(),
hb_fsCloseProcess()
+ added .prg functions HB_OPENPROCESS(), HB_PROCESSVALUE(),
HB_CLOSEPROCESS()
Based on xHarbour code by Giancarlo Niccolai.
Warning: it's possible that they will be changed in the future.
Please test current implementation. Now few notes about it.
Each handle returned by HB_OPENPROCESS() have to be closed
by HB_PROCESSVALUE(). Even if process is killed by HB_CLOSEPROCESS()
its handle is still open and HB_PROCESSVALUE() has to be executed.
Type of handle depends on OS. In *nixes it's process PID. In MS-Windows
it's HANDLE. HB_PROCESSVALUE() attach process exit result in *nixes
cleaning zombie processes and in Windows closing the HANDLE. If you
do not call this function then you will have resource leak.
HB_CLOSEPROCESS() only sends quite request to the process but does
not execute any cleanup code.
All communication handles returned in parameters by reference by
HB_OPENPROCESS() function have to be closed using FCLOSE() function.
If hStdOut and hStdErr are references to the same variables then
executing process stdout and stderr is redirected to the only one
pipe. When handles for stdout, stderr and stdin are not given then
executed process inherits them from parent process unless <lDetach>
parameter is not given. In such case they are redirected to null
device (/dev/null or NUL).
The OS2 version is not tested. Please make tests.
There is no support for DOS which is single process OS.
In WinCE builds std{out,err,in} redirecting and process detaching does
not work too.
The parameters parsing should be updated. Now MS-Windows version uses
native OS command line parsing and quoting by "" can be used for
parameters with blank characters. I do not know any escape character
which can be used to pass (") as parameter to MS-Windows application.
In all other OS-es standard bourne shell rules are used. Parameters can
be quoted by "" or '' and escape character is \. Quoting by '' disables
special meaning of escape character. In OS2 where \ is path separator
escaping and '' quoting is disabled so it works like in MS-Windows
but I do not know if parameters divided by calling process are fully
respected by low level API calls in this system.
* harbour/source/rtl/cdpapi.c
* harbour/source/rtl/langapi.c
* increase maximum number of lang and codepage modules from 64 to 128
(we already have more then 64)
* harbour/include/hbgtinfo.ch
+ added HB_GTI_BOXCP - it allows to chose CP for given characters put
by DISPBOX() and similar box drawing functions.
* harbour/source/rtl/gttrm/gttrm.c
+ added support for HB_GTI_BOXCP. Now it's possible to change the CP
used for box drawing characters. Before this modification is was
hardcoded to "EN"
* use codepage set by HB_SET{DISP|TERM}CP() also for box characters
if last parameter <lBox> is true
This functionality (using different CP for box drawing characters)
is available only if terminal works in Unicode mode (UTF8).
* harbour/source/vm/fm.c
* harbour/include/hbwmain.c
* use HeapAlloc( GetProcessHeap(), ... ) instead of LocalAlloc()
when HB_FM_WIN32_ALLOC macro is set - Toninho's tests show that
it's a little bit more efficient - thanks.
* harbour/contrib/hbct/ftoc.c
* harbour/contrib/hbct/misc1.c
* changed XTOC(), FTOC() and CTOF() to always store/retirve numbers as
double value in little endian order to reduce platform differences
; TOFIX intorduce new macro which can be used instead sizeof( double )
for platforms where double size is different then 8 and macro
HB_FORCE_IEEE754_DOUBLE is set
* harbour/contrib/hbct/dattime2.c
* changed date functions to use default date like in CT3
* harbour/source/vm/classes.c
* updated __objHasMsgAssigned()
+ accept also symbols (@methodname()) not only strings as message name
(2-nd parameter) just like in __objHasMsg() or __objSendMsg()
+ added suport for scalar classes in __objHasMsgAssigned()
+ generate RT error when wrong parameters are passed
* updated comments with .prg functions syntax
* harbour/config/os2/global.cf
* simplified and fixed an error in deleting files
* harbour/source/rtl/hbffind.c
! fixed hb_fsFindNextLow() in OS/2 which was using stat to find file size, but
as it seems, st_size contains garbage. Removed call to stat() and used
buffer returned by DosFindFirst().
* harbour/source/rtl/hbntos.c
* harbour/source/rtl/hbrandom.c
* removed dummy spaces
* harbour/source/rtl/filesys.c
* updated HB_TRACE messages
* harbour/source/rdd/dbfntx/dbfntx1.c
* harbour/source/rdd/dbfnsx/dbfnsx1.c
* return pure index name without path and extension for DBOI_BAGNAME just
like DBFCDX does. It's not Clipper compatible but looks that people
expects ordBagName() behavior implemented in COMIX CDX driver (CL53).
Full index name can be accessed by DBOI_FULLPATH.
If you think it's wrong and we should keep CL52 DBFNTX behavior then
I can revert it but I will want to also change DBFCDX to keep
ordBagName() compatible core RDDs.
Please let me know your preferences.
* harbour/contrib/hbtpathy/telepath.prg
! replaced ThreadSleep() with hb_idleSleep()
* source/vm/classes.c
+ addded __objHasMsgAssigned( object, "msgName" )
; this is similar to xHarbour __clsMsgAssigned() and checks
; if a VIRTUAL message has been overrided in a sub-class.
; TODO: Please check this implementation, I'm sure it can be optimized.
* harbour/source/rdd/dbsql.c
* harbour/source/rdd/dbfntx/dbfntx1.c
* harbour/source/rdd/dbfnsx/dbfnsx1.c
* harbour/source/rdd/dbfcdx/dbfcdx1.c
* harbour/contrib/rddads/ads1.c
! reuse the same error object when CANRETRY flag is set for valid
TRIES member update
* harbour/include/hbgtcore.h
* harbour/source/rtl/hbgtcore.c
* changed BYTE bClearColor to USHORT uiClearColor
The following modifications were made by Viktor Szakats,
(2008-11-22 16:15 UTC+0100 - 2008-12-19 12:56 UTC+0100)
Many thanks. It's the first part without CP conversions in OS calls which
I will want to check yet. Please tests MS-Windows only modifications.
* harbour/make_vc.mak
* harbour/contrib/mtpl_vc.mak
+ Added support for A_USR.
* harbour/include/set.ch
* harbour/include/hbextern.ch
* harbour/source/vm/cmdarg.c
* harbour/source/vm/set.c
* harbour/source/rtl/tgetlist.prg
* harbour/source/rtl/gete.c
* harbour/contrib/hbhpdf/harupdf.c
* formatting
* harbour/source/rtl/hbffind.c
* Changed few places to use sizeof() instead of
explicit size macro.
* harbour/source/vm/extrap.c
+ Added module listing. (also supports Win64)
* harbour/source/common/hbver.c
+ Added Intel C compiler detection. (from xhb)
+ Added Wine detection (unofficial but stable looking
detection method. There intentionally doesn't exist
an official method to do this detection, yet there
are cases when this is necessary. It's also essential
for error reports.)
See also:
http://www.mail-archive.com/wine-devel@winehq.org/msg48659.html
* harbour/source/common/hbfopen.c
! Reverted recent change to use fopen_s(). fopen_s()
will always open in exclusive mode, so it could create
incompatibilities.
* harbour/source/codepage/cpbgmik.c
! Fixed BGMIK collation encoding. Someone familiar with
BG CPs please confirm this, but previously it was
almost certainly broken, as the collation was identical
to BG866.
* harbour/source/rtl/diskspac.c
* harbour/source/rtl/disksphb.c
! GetModuleHandleA() -> GetModuleHandle()
; NOTE: GetModuleHandleA() WinCE emulation in
source/common/wince.c might be unnecessary
after this change.
* harbour/source/rtl/memoedit.prg
* harbour/source/rtl/teditor.prg
! Cleaned insert handling. It now won't anymore mess up
the cursor on instantiation (in New() method). Internal
:lInsert state replaced by Set( _SET_INSERT ). Cursor
is only touched in the editing loop.
* harbour/source/rtl/version.c
! HB_VERSION( HB_VERSION_BUILD_DATE ): Fixed month
being off by one.
* harbour/contrib/hbmzip/hbmzip.c
! Fixed typo which caused timestamps stored in .zip files
to be wrong.
* harbour/contrib/hbhpdf/tests/harupdf.prg
! Minor typo in code.
* harbour/contrib/rddads/rddads.h
+ Added detection for ACE 9.10.
* harbour/contrib/hbtip/sendmail.prg
! Stricter parameter checking in hb_SendMail().
* harbour/contrib/examples/dbu/hb_dbu.dif
+ Added SET DATE ANSI and SET CENTURY ON
* harbour/source/rtl/gtxwc/gtxwc.c
+ Generating HB_K_RESIZE keystrokes on screen resize.
This way it's working similarly to gtwvt.
! Typos in comments.
* harbour/source/rtl/gtwin/gtwin.c
! Fixed HB_GTI_KBDSHIFTS not working in console mode.
Replaced GetKeyboardState() call with GetKeyState() calls.
* harbour/source/rtl/gtwvt/gtwvt.c
+ Added 'layered' window support. It's enabled on W2K
and above (and only when not in Terminal Services
environment, because it makes it perform worse). This
way the screen is also properly displayed regardless
of what the Harbour application does. It also fixes
numerous paint artifacts, when GTWVT window was in
the background and another window is moved in front
of it.
! Fixed only generate HB_K_RESIZE keystrokes when
resize mode is HB_GTI_RESIZEMODE_ROWS.
* harbour/include/hbexprb.c
+ added optimizations for var := var <op> <exp> => var <op>= <exp>
when <op> is +, -, *, / %, ^ operator
% cleanup up checking if message can be used as LValue
* harbour/include/hbcomp.h
* harbour/include/hbmacro.ch
* harbour/source/compiler/ppcomp.c
* harbour/source/compiler/hbusage.c
* harbour/source/compiler/cmdcheck.c
+ added -ko (allow operator optimizations) compiler switch
It's disabled by default.
* harbour/include/hbexprb.c
* removed HB_ADD_SUB_ONE_OPT macro and enable optimizations covered
by it when -ko switch is used.
* harbour/source/common/expropt2.c
* disabled <exp> + 0 => <exp>, <exp> + "" => <exp>, - - <exp> => <exp>
optimizations for default compiler settings. They can be enabled by
-ko compiler switch
+ added optimizations for numeric values and + operator which uses
standard mathematical + behavior for real numbers:
a + b + c == a + ( b + c )
a + b == b + a
It's disabled by default and can be enabled by -ko compiler switch.
There are also other similar optimizations which can be added in
this way.
* harbour/include/hbapi.h
* harbour/source/vm/itemapi.c
* harbour/source/vm/fm.c
! fixed redundant call to hb_xrealloc() in string resize operation.
Thanks to Mindaugas for locating it.
* harbour/source/rdd/dbfnsx/dbfnsx1.c
* added some missing const in char * declarations
* harbour/contrib/hbtip/sendmail.prg
* cleaned unused assigned detected by new Mindaugas logic
* harbour/bin/hb-func.sh
+ added support for rddsql and sdd* libraries in hb* scripts
* harbour/harbour-win-spec
* harbour/harbour-wce-spec
! added rddnsx library
* harbour/mpkg_rpm.sh
* harbour/harbour.spec
+ added harbour-fbird RPM and support for sqlrdd and sdd* libraries
* harbour/contrib/rddsql/sddpg/Makefile
+ added other paths checking for postgresql header files
* harbour/contrib/rddsql/sddfb/Makefile
! fixed file name
* harbour/contrib/rddsql/sddpg/pgsqldd.c
* harbour/contrib/rddsql/sddfb/fbirddd.c
* cleanup init symbol declaration
! casting for C++ compilation
* added missing const in some char * declarations
* harbour/contrib/rddsql/sddfb/fbirddd.c
* changed some parameters passed to firebird functions.
Mindaugas please verify me because these modifications changes
the code behavior. I do not know FireBird API and I modified it to
only pass C++ compilation checking error messages. At least these
lines should be checked: 233, 236, 244, 245, 260, 261, 440, 445
If possible please also check casting I added in in other lines
because it may hide some mistakes.
* source/compiler/hbopt.c
+ added logic to generate 'assigned but unused' warning in case
op= is used. Ex.:
nI += nJ
nI := 5
; This type of redundancy is not found in core code
* harbour/include/hbstack.h
* harbour/include/hbapi.h
* harbour/include/hbapiitm.h
* harbour/source/vm/hashes.c
* small modifications in order of some structure members for better
alignment
* harbour/source/vm/hvm.c
* pacified BCC warnings
* harbour/contrib/rddsql/Makefile
* harbour/contrib/rddsql/sddpg/Makefile
+ harbour/contrib/rddsql/sddfb/Makefile
* updated GNU make files for new SQL RDDs - many thanks to
Mindaugas
* contrib/gtwvg/wvgclass.prg
* contrib/hbgd/gdchart.prg
! fixed assigned but not used variable
; Also I would like to wish you all a happy Christmas holidays
* include/hbextern.ch
* source/rtl/seconds.c
+ implemented hb_milliseconds(). This function returns time value
from some moment in the past (not midnigth!). It does not start
to cound from zero in midnight, thus, can be safetly used to
measure time intervals.
+ tests/ticktime.c
+ Added nice test for hb_milliseconds().
Results on WinXP:
Ticks per second: 63.595
Min/avg/max interval (ms): 15.000 / 15.724 / 47.000
Loops per tick: 13035.09
Results on openSUSE on VirtualBox on WinXP:
Ticks per second: 384306.138
Min/avg/max interval (ms): 0.001 / 0.003 / 11.112
Loops per tick: 1.04
* contrib/rddsql/sqlbase.c
* contrib/rddsql/sqlmix.c
* contrib/rddsql/tests/test1.prg
* contrib/rddsql/common.mak
- contrib/rddsql/hbsqldd.h
+ contrib/rddsql/hbrddsql.h
* contrib/rddsql/make_b32.bat
* contrib/rddsql/make_vc.bat
* contrib/rddsql/make_gcc.sh
+ contrib/rddsql/sddmy
- contrib/rddsql/mysqldd.c
+ contrib/rddsql/sddmy/mysqldd.c
+ contrib/rddsql/sddmy/Makefile
+ contrib/rddsql/sddmy/make_gcc.sh
+ contrib/rddsql/sddmy/common.mak
+ contrib/rddsql/sddmy/make_b32.bat
+ contrib/rddsql/sddmy/make_vc.bat
* changed structure and names of rddsql libraries. Since it can
support many SQL databases, each backend moved to a separate
library. Otherwise (all backend in the same library) nobody will
be able to compile library without installing packages of ALL
supported SQL databases. Now:
rddsql - RDD for interfacing to SQL database drivers
sddmy - MySQL database driver
sddpg - Postgre SQL database driver
...
; TOFIX: I was unable to manage build files.
contrib/sqlrdd/sddmy/make_b32.bat sets HB_ROOT variable:
set HB_ROOT = ..\..\..
but this variable is lost and contrib/make_b32.mak reassign it:
!ifndef HB_ROOT
HB_ROOT = ..\..
!endif
This breaks building of library. To fix this problem, I've added
a two hacks:
- added additional include path into CFLAGS (see make_b32.bat)
- added "..\" to LIB_PATH (see common.mak)
The same problem is for both make_b32 and make_vc, and for both
sddmy and sddpg libraries.
; TODO: check (and adjust if neccessary) build files for unix
+ contrib/rddsql/sddpg
+ contrib/rddsql/sddpg/pgsqldd.c
+ contrib/rddsql/sddpg/Makefile
+ contrib/rddsql/sddpg/make_gcc.sh
+ contrib/rddsql/sddpg/common.mak
+ contrib/rddsql/sddpg/make_b32.bat
+ contrib/rddsql/sddpg/make_vc.bat
+ added SQL database driver for Postgre SQL
; does anyone have more RDD for Christmas gift? :)
* harbour.spec
+ added rddnsx to rpm builds. Thanks Przemek.
; Also I would like to wish you all a happy xmas holidays
; and a new year full of health, love and Harbour business :)
+ harbour/include/hbrddnsx.h
+ harbour/source/rdd/dbfnsx
+ harbour/source/rdd/dbfnsx/dbfnsx1.c
+ harbour/source/rdd/dbfnsx/Makefile
* harbour/config/global.cf
* harbour/make_b32.mak
* harbour/make_vc.mak
* harbour/make_gcc.mak
* harbour/common.mak
+ added new DBFNSX RDD. Those of you who knows SIX3 should also
know NSX index format. It's very interesting format and in some
cases much better then CDX. I hope you will find it interesting.
In hbrddnsx.h you can find detail information about Harbour NSX
implementation.
It's a small gift from me for Christmas ;-)
I would like to wish all of you Marry Christmas and Happy new Year.
best regards,
Przemek
* include/hberrors.h
* source/compiler/hbgenerr.c
* source/compiler/hbopt.c
+ implemented warning: Variable '%s' is never assigned in function %s(%d).
This variable can be replaced by value NIL
* source/rtl/treport.prg
% removed never assigned variable
* harbour/source/common/expropt1.c
* harbour/source/common/expropt2.c
! fixed wrongly modified expression during optimization for
( <numConst> / 0.0 )
* cleaned bDec/bWidth updating in compile time optimizations
* harbour/tests/speedtst.prg
! fixed possible race condition in initialization
+ added set workarea private for xHarbour builds - thanks to Phil
* harbour/doc/cmpopt.txt
* small cleanup
* harbour/include/hbvmpub.h
* harbour/source/vm/hvm.c
% some optimizations mostly for MT mode or compilers which do
not support function autoinline optimization
* harbour/include/hbsetup.h
* temporary disabled noreturn function attributes - we have some
code which puts references to some functions after hb_errInternal()
call to force link bindings which is fully optimized if hb_errInternal
is declared with noreturn attribute so we cannot use it as long as we
will not clean them.
* harbour/tests/speedtst.prg
* build default xHarbour binaries without support.
xHarbour needs separated version for MT and ST mode
because standard MT functions are not available in
ST libraries.
* harbour/ChangeLog
! cleaned some typos in my recent ChangeLog entries
* harbour/include/hbvm.h
* harbour/include/hbvmpub.h
* harbour/include/hbsetup.h
* harbour/include/hbapi.h
* harbour/include/hbapierr.h
* added some function attributes which can improve code optimization
or introduce some additional warnings.
Now it works only in GCC builds.
* harbour/source/vm/hvm.c
! fixed missing double decimal places initialization in some number
negating
! fixed setting number of decimal places in plus and minus operations
if integer overflow forces conversion to double item
* some minor optimizations
* harbour/source/rtl/rat.c
* formatting
* harbour/contrib/gtwvg/gtwvg.h
! Seperated PHB_GTWVT structure for GUI elements as PHB_GUIDATA + PHB_GTWVT.
* harbour/contrib/gtwvg/gtwvg.c
* harbour/contrib/gtwvg/wvgcore.c
* harbour/contrib/gtwvg/wvgutils.c
! Modified to respect PHB_GUIDATA structure.
* harbour/contrib/gtwvg/wvgpaint.prg
! Modified THREAD STATIC paint_ behavior to respect true thread oriented
until THREAD STATIC is modified to respect its true meaning.
* harbour/contrib/gtwvg/wincback.prg
* harbour/contrib/gtwvg/wvgmenub.prg
* harbour/contrib/gtwvg/wvgclass.prg
* harbour/contrib/gtwvg/wvgsle.prg
* harbour/contrib/gtwvg/wvgstatc.prg
* harbour/contrib/gtwvg/wvgsysw.prg
* harbour/contrib/gtwvg/wvgwnd.prg
! Fixed to respect new compiler warnings.
Thanks Mindaugus for this great addition.
Hope code will be faster than before.
