* harbour/src/rdd/dbf1.c
+ allow to open DBF files with 0x07 and 0x87 signature created by
CA-VO DBFNTX when ANSI encoding is enabled.
Thanks to Vitomir Cvitanovic for CA-VO tests.
Warning! After any modifications in such files Harbour DBF* RDDs
will correct above signatures to standard DBASE ones (0x03 and 0x83).
* utils/hbmk2/hbmk2.prg
+ Quite big commit aiming to clean path forming in different
situations. After this is change it should be more or less
true that internally hbmk2 always uses native path format,
it should accept _any_ format regardless of platform,
and it should convert filenames to proper format only when
calling external compiler tools, according to their specific
taste.
F.e. this should fix problems when doing Windows cross-build
on *nix systems, with watcom, or other compilers (except
mingw) which has native compiler. It should continue to
work with WINE based on compilers in similar situations.
There is one specific exception when passing around
filenames as part of options. In this case hbmk2 cannot do
too much magic, so try to avoid it.
Though I never actualluy tried this latter case.
Please test it, regressions are possible. I'd be happy to
see stress tests with passing various combination of pathseps
and watching -trace output whether everything is properly
converted to right format.
+ Will now convert filenames to Cygwin format for Cygwin
targets.
! Fixed to accept .hbi files without '@' prefix.
* config/postinst.prg
% Do not add '@' prefix when referring to .hbi files.
* src/vm/extend.c
! Fixed TRACE message of new hb_parnintdef() function.
* src/rtl/hbcomhb.c
* Minor correction to one function description in comment.
* src/rtl/hbsocket.c
* Minor formatting.
* include/hbcom.ch
+ Added comments.
* Formatting.
+ Added comment that code is used by .c code.
* contrib/hbtpathy/telepath.prg
* Rewritten to use native HB_COM*() API (instead of hbct
specific COM_() one).
+ tp_send() code rewritten to use core timeout functionality.
+ Implemented formerly disabled tp_ctrldtr() function.
(pls review it)
% FetchChars() internal function simplified.
* contrib/hbcomm/hbcomm.prg
* Rewritten to use native HB_COM*() API (instead of hbct
specific COM_() one).
* contrib/hbsms/hbsms.prg
* Rewritten to use native HB_COM*() API (instead of hbct
specific COM_() one).
* Receive code rewritten to rely on core timeout functionality
instead of rolling local implementation.
; I didn't make any tests with API converted COM code, so please
review and test these changes.
* contrib/hbtpathy/hbtpathy.hbc
* contrib/hbcomm/hbcomm.hbc
* contrib/hbsms/hbsms.hbc
- Deleted hbct dependency.
* INSTALL
* package/winuni/RELNOTES
* QT 4.6.2 -> 4.6.3
* harbour/src/rtl/hbsockhb.c
* changed return values of some functions: nSuccess to lSuccess
* HB_SOCKET_PF_* changed to HB_SOCKET_AF_*
- hb_parnintdef() (moved to extend.c)
! some bug fixes
! some typo in documentation
* harbour/examples/udpds/udpds.prg
* sychronized with hb_socket*() changes
* changed server thread exit condition (proposed by Przemek)
* harbour/include/hbapi.h
* harbour/src/vm/extend.c
+ hb_parnintdef()
* harbour/src/rtl/hbsocket.c
* harbour/include/hbsocket.ch
! cleaned a little bit HB_SOCKET_AF_*/HB_SOCKET_PF_* usage to not confuse
other developers and users.
* src/rtl/hbregexc.c
+ Restored RTE which is thrown when unsupported regex string is
found in any .prg level regex API calls except HB_REGEXCOMP().
The only difference compared to previous solution is different
error code (3015) to differenciate this case from wrong
parameter _type_ error cases.
; TOFIX: IMO even current solution is not ideal, because it just
forces developers to use HB_ISREGEX() to pre-evaluate regex
strings before actually using them (causing performance hit),
or it forces them to use HB_REGEXCOMP() to pre-compile and
pre-evaluate the expressions before passing them to actual
regex functions, in the valid case they want to avoid unexpected
RTEs due to missing or unsupported regex (PCRE) subsystem.
IOW it makes plain HB_ATX( <cRegExp>, ... ) style calls
prone to fail with RTE in certain, very hard to predict
situations in officially supported Harbour builds.
Comments are appreciated.
* utils/hbmk2/hbmk2.prg
* Deleted unused code.
* Formatting.
* harbour/src/rtl/hbsocket.c
! added missing code to duplicate IP6 address passed explicitly
to hb_socketResolveAddr() - thanks to Mindaugas for reporting
the problem
* src/rtl/hbregexc.c
* Changed to not throw RTE, just gracefully return
error value when invalid PCRE expression is used.
It will continue to throw RTE only for user-controllable
error situations, f.e. when wrong argument type is passed.
! Fixed to not contain English text hard-coded in RTE message.
* src/rtl/hbregex.c
* HB_ATX() changed to fill parameters passed by reference with
error values (zeros) in case of error.
* HB_REGEX(), HB_REGEXSPLIT(), HB_REGEXATX(), HB_REGEXALL()
functions changed to consistently return arrays. If there
is no match, the array will be empty. This gives a more
natural (and Clipper/Harbour-like) interface (f.e. return
value can be passed to FOR EACH loop without extra checks
and extra local variable).
Slight INCOMPATIBILITY for those who did 'r == NIL' check
to verify non-match and something special in this case.
For code simply evaluating the returned matches, the NIL
check can be deleted, and for code which wants to be
backward/forward compatible _and_ to specifically check
for no match situation it's suggested to use 'EMPTY( r )'.
Please check me.
* HB_REGEXMATCH() marked as HB_LEGACY_LEVEL4. (this leaves
quite some headstart to change it)
* include/hbextern.ch
- Deleted HB_REGEXMATCH().
* utils/hbmk2/hbmk2.prg
% Simplified regex usage after above changes.
+ Will now show an internal error in case the used
regex is invalid in current Harbour context
(f.e. when no PCRE engine was forced by user, or
the one used doesn't support the expression used by
hbmk2). I don't know how special current expression
is (it doesn't seem very much so), but it can be
streamlined to something more compatible if current
is deemed not ideal.
* src/compiler/hbusage.c
* Changed to enclose e-mails between '()' instead of '<>'.
* include/hbsetup.h
! Patch by Tamas Tevesz for SunPro compiler.
* ChangeLog
- Deleted my personal exception from ChangeLog license.
* src/rtl/gtgui/gtgui.c
* src/rtl/gtwin/gtwin.c
* src/rtl/gtwvt/gtwvt.c
* contrib/gtwvg/gtwvg.c
* contrib/gtwvg/wvggui.c
* examples/gtwvw/gtwvw.c
+ Will now put to clipboard and request from clipboard UNICODE
text in UNICODE (default) mode. This should fix CP conversion
problems experienced in non-UNICODE mode (more in a separate
bug report).
Please review me.
* src/rtl/memvarhb.prg
* Minor in comment.
* contrib/hbtpathy/tpunix.c
+ Applied patch sent to the list by Tamas Tevesz.
Fixing some compilation problems due to missing predefined macros
on certain *nix systems.
* contrib/hbqt/hbqt_common.hbm
* Formatting.
* utils/hbmk2/hbmk2.prg
! Fixed to make sure -depcontrol= header path value is added
to the header path search list in the same location as specified
on the cmdline.
* utils/hbmk2/hbmk2.prg
+ Now uses regular expression to extract #include directives
from source files, so now it should recognize all supported
syntax variations (f.e. " # include <>"). The only
difference between real source parser is now only the lack
of comment filtering (and no preprocessor, but this latter
isn't desired for multiple reasons).
Pls review me.
* src/rtl/hbregex.c
* HB_ATX() cleanup code layout, minimalized variable scopes.
% HB_ATX() minor optimization: No need to check pcount before
calling hb_stor*() functions since it's done anyway inside
those calls.
* HB_ATX() will now default the 5th <nEnd> parameter to the
full length of the passed source string. It's slight
incompatibility though I doubt too many users would have
called this with empty 5th parameter, since the function
just didn't do anything in this case.
Pls review this, maybe I misunderstood original intent.
* harbour/src/debug/dbgentry.c
* harbour/src/vm/hvm.c
+ added support for breakpoints and file wide static variables
when debugging non main thread.
It finished modifications in debugger code for MT mode.
Probably we should add yet to debugger support for separate
GT window(s) with debugger data for ST and MT applications.
* harbour/src/vm/garbage.c
+ added new PRG function HB_GCSETAUTO( [<nNewVal>] ) -> <nPrevVal>
which allows to set the frequency of automatic GC activation.
<nNewVal> is number of GC items (in thousands). If total number
of new GC items allocated from last GC pass exceed this value
then GC is automatically activated. Setting 0 disable automatic
GC activation. Automatic GC activation is necessary for programs
which allocates big number of complex items with cyclic references
without entering any idle state which can activate GC. In such
case users have to add to their code call to hb_gcAll() function.
In xHarbour they used sometimes background task for this job
but it's inefficient workaround for the problem due to reduced
performance.
This code is still experimental code enabled by HB_GC_AUTO Harbour
compile time macro.
* harbour/src/debug/dbgentry.c
! fixed potential memory leak in unused code for GLOBAL [EXTERN]
variables
! cleaned a little bit code for line numbers with good break points
* src/vm/fm.c
* src/vm/set.c
* SET( _SET_HBOUTLOG, NIL ) will now reset the internal output
name to NULL, which is also the new internal default value.
* After this change the GPF handler and FM STAT dump code
will do the defaulting to "hb_out.log" filename, if the
internal _SET_HBOUTLOG value is NULL. This causes a small
incompatibility in how log file can be disabled, and at the
same time allows to implement low-level logging logic for
default case which is safer than current one, while allowing
the user (developer) to set a custom log filename and expect
regular log file append behavior.
From now on to disable logging, use: SET( _SET_HBOUTLOG )
* utils/hbmk2/hbmk2.prg
! Fixed to not do pkg detection and key header evaluation
after -stop and certain modes.
+ Added full list of std C and POSIX headers to the header
exclusion list in header dependency parser code.
Also added dos.h.
* contrib/xhb/dbgfxc.c
* Formatting.
* contrib/xhb/hbcompat.ch
* contrib/xhb/xhbfs.c
! Deleted FILESIZE(). It's not an xhb compatibility function,
but one offered as part of CT lib in xhb, and HBCT in Harbour.
* src/common/hbver.c
+ Added compiler version detection for clang 2.x (and newer)
Patch by Tamas Tevesz.
* INSTALL
* Minor.
* config/bsd/clang.mk
* config/linux/clang.mk
+ Clearing forced C++ mode if clang 1.x is detected.
(Patch from Tamas Tevesz. Slight fix added by me to
set HB_CMP when falling back to C mode. I didn't make
tests though.)
; NOTE: Probably HB_BUILD_MODE=c should be export-ed
to avoid double evaluation. Pls test it.
* config/postinst.prg
* config/ren_sfn.prg
+ Raised warning level to 3.
* src/common/hbver.c
+ Fine tuned SunPro version detection.
Patch submitted by Tamas Tevesz.
* contrib/hbwin/win_shell.c
* contrib/hbwin/tests/testcopy.prg
! WIN_SHFILEOPERATION() fixed after initial upload.
+ Added more test code.
* utils/hbmk2/hbmk2.prg
* '-plug=' option renamed to '-plugin='
+ Added support for 'plugins=' line in .hbc files.
% Plugins are now fully loaded just once at the beginning
of hbmk2. (as opposed to every invocation)
+ Plugins are now automatically treated as .hrb or .prg
based on _file content_. This means that any extension
can be used for plugins for both .prg and .hrb code.
When .prg or .hrb extension is used there isn't any
extra trial made on the file content, it will be load
as source or HRB respectively.
Maybe we should find a new distinctive extension for
hbmk2 plugins.
* Default extension for -plugin= option changed to .prg
(was: .hrb)
+ Showing type of input plugin in -trace mode.
('source' or 'compiled')
* config/detect.mk
! Applied fix to DragonFly patch, submitted by Tamas Tevesz.
* src/vm/runner.c
* Minor formatting.
* src/common/hbprintf.c
* include/hbsetup.h
* config/global.mk
* config/bsd/libs.mk
* config/detect.mk
+ Applied patch by Tamas Tevesz, making it possible to build
Harbour for DragonFly BSD systems.
Thanks a bunch.
* utils/hbmk2/hbmk2.prg
+ Applied above change to hbmk2 code, so now it should also
support DragonFly BSD. (pls test it)
* harbour/include/hbapi.h
* harbour/src/vm/hashes.c
* harbour/src/vm/hashfunc.c
+ added support for keeping strict assign order in hash arrays. It's
enabled optionally by setting HB_HASH_KEEPORDER hash array flag.
It gives the same functionality as associative arrays in xHarbour
(enabled by HSETAACOMPATIBILITY()) but this implementation is
internally completely different. It does not introduce linear
scan in add operation so enabling it should not reduce the
performance in this operation. It can even improve it on some
hardware reducing number of memory bytes which have to be moved.
Only delete operation will force linear index scan. The most
important in this implementation is that it does not need any
additional functions like HAA*() in xHarbour. Just simply all
existing functions operating on position indexes like HB_HPOS(),
HB_HKEYAT(), HB_HVALUEAT(), HB_HPAIRAT(), HB_HDELAT(), HB_HSCAN()
use as index natural order in which items were added to hash array.
Also HB_HKEYS(), HB_HVALUES() and FOR EACH iterations respect it.
+ added new PRG functions
HB_HKEEPORDER( <hValue> [, <lNewSetting> ] ) -> <lPrevSetting>
HB_HSETORDER( <hValue> [, <lNewSetting> ] ) -> <hValue>
which cam be used to enable/disable strict order in hash array.
Enabling strict order for non empty hash arrays accept the order
created after sorting existing item not the original assign order.
Disabling strict order for non empty hash arrays may change the
items order.
; TODO: add translation for xHarbour's HAA*() functions to hbcompat.ch
and/or xhb library.
* harbour/contrib/hbnetio/netiocli.c
* harbour/contrib/hbnetio/netiosrv.c
% reenabled TCP_NODELAY on client and server side
I had to be really tired when I was making tests and I mixed
hb_socketSetNoDelay() with hb_socketSetBlockingIO().
Thanks to Aleksander Czajczynski who noticed the delay in his
tests.
* harbour/contrib/hbwin/olecore.c
* minor cleanup
* harbour/src/vm/hashes.c
* do not clone preallocated unused area
* harbour/contrib/hbnetio/netiocli.c
* disabled RT errors inside NETIO_CONNECT() and NETIO_GETCONNECTION()
when wrong password or NETIO server is used but return .F.
* harbour/include/hbapi.h
* harbour/include/hbvmint.h
* harbour/include/hbvmopt.h
* harbour/include/hbstack.h
* harbour/include/hbvmpub.h
* harbour/include/hbtypes.h
* harbour/src/vm/estack.c
* harbour/src/vm/hvmall.c
* slightly modified the names of HB_{API|STACK}_MACROS.
These are internal macros in current Harbour code but they are
still defined in some old user build scripts used to compiled
user or core code what could cause unpredictable results for
final binaries. AFAIR default xMate configuration enables them.
* harbour/src/rdd/dbf1.c
* minor cleanup
* harbour/include/hbapi.h
+ added new macro HB_IS_EVALITEM() - returns true for items which can
understand EVAL message i.e. CODEBLOCK and SYMBOL items.
* harbour/contrib/hbwin/axcore.c
* harbour/contrib/hbwin/olecore.c
* harbour/contrib/hbwin/hbwinole.h
* modified hb_oleDispInvoke() function to accept as additional
parameter pointer to HVM item instead of pointer to DISPID.
Such version is more universal.
* harbour/include/hbapi.h
* harbour/src/vm/hashes.c
+ added new C function hb_hashGetCItemPos()
* harbour/src/rtl/hbcom.c
! fixed typo located by Bisz István
* src/rtl/memvarhb.prg
* HB_MVSAVE(): reset to be a FUNCTION to avoid the "volatile"
return value and better imitate __MVSAVE() behavior.
% HB_MVSAVE(): deleted unnecessary UPPER().
- Deleted TODO.
* ChangeLog
* Marked TODOs as done. (Thanks Przemek)
* harbour/include/hbvm.h
* harbour/src/vm/hvm.c
+ added new C function hb_vmIsActive()
* harbour/src/rtl/memvarhb.prg
! fixed procedure returning value
! store only simple variables in HB_MVSAVE() just like __MSAVE()
Possible extension: add option to store also NIL, arrays, hashes
and maybe object variables.
! restore variables in HB_MVRESTORE() as private ones just like
__MVRESTORE()
* harbour/contrib/hbwin/oleinit.c
* harbour/contrib/hbwin/olecore.c
* updated to compile and link in WinCE builds. Now code using OLE
can be linked with MinGWCE builds without any errors but I have
no idea if it works. Can someone check it?
* utils/hbmk2/hbmk2.pt_BR.po
* utils/hbmk2/hbmk2.hu_HU.po
* utils/hbmk2/hbmk2.prg
+ Added experimental -hbdynvm mode.
+ Added support for .def input file in -hbdyn/-hbdynvm modes.
; TODO: Couldn't find how to pass .def files to Open Watcom wlink.
Anybody with an experience here?
* src/rtl/memvarhb.prg
+ Added one initialization to NIL.
* include/hbextern.ch
* src/rtl/Makefile
+ src/rtl/memvarhb.prg
+ Added:
HB_MVSAVE( <cFileName>, [<cMask>], [<lIncludeMask>] )
HB_MVRESTORE( <cFileName>, [<lAdditive>], [<cMask>], [<lIncludeMask>] )
They serve the same purpose with same interface
and similar behavior to __MVSAVE()/__MVRESTORE()
(and __MSAVE()/__MRESTORE(), and 'SAVE TO' command),
only that they support longer than 10 char variable names.
It is not possible to add such support in Clipper-style
.mem files, so these functions will create an incompatible
file format with default '.hbv' extension.
; TOFIX: __MVRESTORE() restores variables as PRIVATEs
while HB_MVRESTORE() as PUBLIC.
; NOTE: GETLIST exclusion is solved with simple trick,
as I couldn't find how it's solved in __MVSAVE().
; Consider this experimental solution to the problem.
* contrib/hbmysql/tmysql.prg
% STOD( "" ) -> STOD()
* contrib/hbide/ideprojmanager.prg
! -hbexec -> -hbexe
* src/rtl/typefile.prg
! __TYPEFILE(): Fixed both RTEs to be Clipper compatible.
* harbour/include/hbapi.h
* harbour/src/vm/dynsym.c
+ added to C functions:
int hb_dynsymToNum( PHB_DYNS pDynSym );
PHB_DYNS hb_dynsymFromNum( int iSymNum );
which can be used to make conversion between dynamic symbols and
numbers
* src/rtl/gtwvt/gtwvt.c
- Deleted notifierblock/fireevent support.
Light is green to add threaded screen refresh and event
handling. This could enhance performance and fix the
frozen screen refresh in Harbour app loops.
; INCOMPATIBLE: Switch to HB_INKEY_GTEVENT + HB_K_* solution.
* src/rtl/gtwvt/gtwvt.c
* Changed HB_GTI_CLOSABLE to not cause any visible change on the UI.
Setting HB_GTI_CLOSABLE to .F., will now cause that a HB_K_CLOSE
event is returned to application by INKEY(), and app can decide
whether to ignore it or QUIT.
This also means that the feedback is now not an async event in
HB_GTI_CLOSABLE .F. mode, IOW if the app is in a loop without
reading keys, close button won't work, which is IMO closer to
how apps are supposed to behave. It also won't grey out the
close buttong anymore which is also a unusual thing to do
amongst apps.
; TODO: If this turns out to be good solution, sync this behavior
with HBWIN.
; TODO: Delete old "fireevent" async even handling using notifier
callback codeblock.
* include/inkey.ch
+ Added new special GT events (over existing HB_K_RESIZE):
HB_K_CLOSE
HB_K_GOTFOCUS
HB_K_LOSTFOCUS
+ Added new Harbour specific INKEY() event filter:
HB_INKEY_GTEVENT
* INKEY() codes below marked as compatibility:
K_MMLEFTDOWN, K_MMRIGHTDOWN, K_MMMIDDLEDOWN
* src/rtl/hbgtcore.c
! Fixed to only return HB_K_RESIZE/CLOSE/GOTFOCUS/LOSTFOCUS
events when allowed via HB_INKEY_GTEVENT filter.
; INCOMPATIBILITY: HB_K_RESIZE will now only be returned
by INKEY() if this filter is added.
* src/rtl/gtwvt/gtwvt.c
- Deleted code protected by __HB_GTWVT_GEN_K_MMDOWN_EVENTS guard.
+ Will now generate INKEY() event:
HB_K_LOSTFOCUS
HB_K_GOTFOCUS
and
HB_K_CLOSE.
; TOFIX: HB_K_CLOSE doesn't work yet. Probably some rearrengement
of window close control will be needed.
; The goal is to be able to keep "fireevent" feature in a Clipper/Harbour
and thread friendly manner, so it becomes possible to move OS level
UI event into separate thread.
* src/rtl/tget.prg
! Fixed "@RE" picture mask handling in GETs.
Thanks to Aleksander Czajczynski for reporting the
issue with clear example.
The fix applied is a de-hacked version of the one
sent to the list. What's simply done is 'R' being
disabled when used in conjunction with 'E'.
