* *
* partial sync with the 3.4 fork codebase. These are the things
synces for the most part:
- copyright headers
- grammar/typos in comments and some readmes
- comment/whitespace/decorations
- variable scoping in C files
- DO CASE/SWITCH and some other alternate syntax usage
- minimal amount of human readable text in strings
- minor code updates
- HB_TRACE() void * casts for pointers and few other changes to
avoid C compiler warnings
- various other, minor code cleanups
- only Harbour/C code/headers were touched in src, utils, contrib,
include. No 3rd party code, no make files, and with just a few
exceptions, no 'tests' code was touched.
- certain components were not touched were 3.4 diverged too much
already, like f.e. hbmk2, hbssl, hbcurl, hbexpat
- the goal was that no actual program logic should be altered by
these changes. Except some possible minor exceptions, any such
change is probably a bug in this patch.
It's a massive patch, if you find anything broken after it, please
open an Issue with the details. Build test was done on macOS.
The goal is make it easier to see what actual code/logic was changed
in 3.4 compared to 3.2 and to make patches easier to apply in both
ways.
* contrib/hbexpat/3rd/expat/_hbconf.h
* contrib/hbexpat/3rd/expat/xmlparse.c
! fixed WINCE builds
! fixed OpenWatcom DOS builds
! fixed OpenWatcom OS2 builds
* contrib/hbexpat/3rd/expat/expat.dif
* rediffed
* include/harbour.hbx
* include/hbvm.h
* src/harbour.def
* src/vm/hvm.c
* src/vm/thread.c
+ added new C function
HB_BOOL hb_vmThreadIsMain( void * );
it checks if given or current thread is main HVM thread
+ added new PRG function
hb_threadIsMain( [ <pThID> ] ) -> <lMainHvmThread>
it returns true if given or current thread is main HVM thread
+ added new PRG function
hb_mutexExists( <pMtx> ) -> <lExists>
it returns true if passed <pMtx> parameter is pointer item to
fully functional mutex
* src/rtl/tpersist.prg
! fixed problem with array item deserialization - many thanks
to Peter Rees for exact information about the problem.
* src/rtl/val.c
* minor simplification
* include/hbcomp.h
* include/hbcompdf.h
* include/hberrors.h
* include/hbexprb.c
* src/common/expropt1.c
* src/compiler/hbgenerr.c
* src/compiler/hbmain.c
+ added new macros HB_ET_MACRO_NOLIST and HB_ET_MACRO_NOPARE
% use new HB_ET_MACRO_* macros
+ added new compile time error:
"Code block contains both macro and with object messages ':%s'"
NOTE: -kd compiler switch allows to compile codeblocks with macros
and declared symbols / with object messages
* replaced hb_compErrorCodeblock() with hb_compErrorCodeblockDecl() and
hb_compErrorCodeblockWith()
+ added new C function hb_compPushMacroVar()
* code simplification
; added few comments
* utils/hbtest/rt_math.prg
* extended test code for macro messages and macro <op>=, pre/post ++/--
operations
* src/vm/hvm.c
! protection against executing hb_threadStateNew() during GC pass
inside hb_vmRequestReenterExt()
* src/vm/garbage.c
! add volatile attribute to s_bCollecting variable
* small modification in hb_gcAll() parameter
* src/rtl/errsys.prg
; minor comment cleanup
* doc/xhb-diff.txt
* extended a little bit section STRONG TYPED VARIABLES
* *
% remove brandings and homepage [1] from copyright header. Pass 1 - using script.
[1] nobody has access to it anymore AFAIK - and it's also just
a redirect since long
! update url in copyright header
; this should make the diff between 3.4 and 3.2 easier to manage
* config/global.mk
* set HARBOUR=F:100 during build process in OS2 builds - it increases
maximum number of open handles per process from 20 to 100 so it should
be possible to use -j<n> switch during Harbour compilation - please
test, i.e.:
os2-make -j3
* contrib/xhb/hbcompat.ch
! fixed pp rules for Str(,,, <lStrip> ) translation
* include/hbapifs.h
* src/rtl/filesys.c
+ added new C function:
HB_FHANDLE hb_fsOpenEx( const char * pszFileName, HB_FATTR nAttr,
HB_USHORT uiFlags );
* minor cleanup
* include/hbmemory.ch
* src/vm/fm.c
+ added two new actions for Memory() function:
HB_MEM_STATISTICS - return non 0 value if FM statistic is enabled
HB_MEM_ISLIMIT - return non 0 value if used memory limit is
supported
+ added new PRG function:
__fm_allocLimit( [ <nNewLimit> ] ) -> <nPrevLimit>
It allows to set limit for maximum used memory in some memory
managers, -1 means no limit.
Now it works with DLMALLOC and in ST mode it's limit for memory
allocated from system for whole application. In MT mode when
HB_FM_DLMT_ALLOC (default for DLMALLOC) is enabled it's limit for
single MSPACE (Harbour allocates 16 MSPACES and balance them
between threads to improve scalability) so it's rather per thread
limit. If memory statistic module is enabled during Harbour build
process then this limit works with any memory manager and defines
limit for total memory allocated by Harbour application regardles
of MT or ST mode.
* src/vm/hvm.c
* use HB_MEM_STATISTICS instead of HB_MEM_USEDMAX to check if memory
statistic module is enabled
* src/vm/hvm.c
* allow to use == operator for codeblocks
* utils/hbtest/rt_hvm.prg
* updated for == operator which accepts codeblocks
* utils/hbtest/rt_main.ch
* do not define __CLIPPER__ macro when ITK Clip compiler is used
* contrib/hbbz2/errint.c
* added HB_EXPORT to bz_internal_error() declaration
* src/vm/hvm.c
+ allow to compare symbol items with == operator, i.e.:
? @QOut() == &( "@QOut()" )
* include/hbapiitm.h
* src/vm/itemapi.c
+ added new C function:
hb_itemCompare( PHB_ITEM pItem1, PHB_ITEM pItem2,
HB_BOOL bForceExact, int * piResult );
For compatible types it compares pItem1 with pItem2 setting piResult
to -1, 0 or 1 if pItem1 is <, == or > then pItem2 and returns true
otherwise it returns false and does not touch piResult.
+ allow to compare symbol items in hb_itemEqual()
* include/harbour.hbx
* src/vm/hvm.c
+ added new PRG function:
__RecoverErrorBlock() -> <bError> | NIL
It extracts previous error block pushed on HVM stack
by BEGIN SEQUENCE WITH ...
* src/rdd/dbf1.c
! fixed detecting VFP tables broken in my previous
modification - thanks to Mario H. Sabado for information
about the problem.
* src/vm/hvm.c
! delay locking of main HVM thread stack until HVM is initialized,
it fixes issue #107
! protect access to HB_STACK_TLS_PRELOAD by s_fHVMActive in
hb_vmLockForce()
* include/hbvmpub.h
* src/vm/hvm.c
+ added support for hb_vmAtInit(), hb_vmAtExit() and hb_vmAtQuit()
functions for dynamically loaded libraries
This modification should also fix issue #33
* src/vm/estack.c
! protect access to HB_STACK_TLS_PRELOAD by hb_stack_ready() macro,
it's necessary for MT builds which do not use direct TLS access
but need TLS key initialization, i.e. OS2
* src/vm/hvm.c
! protect access to HB_STACK_TLS_PRELOAD by s_fHVMActive
* include/harbour.hbx
* src/vm/hvm.c
+ added new PRG function:
__BreakBlock() -> {|e| Break( e ) }
* include/hbexprb.c
% replace {|e| Break( e ) } with __BreakBlock() function call.
__BreakBlock() returns exactly the same codeblock on each call
so using it save memory and improve a little bit speed because
it's not necessary to allocate new GC memory block and register
it in GC block list. Additionally in MT mode it eliminates mutex
lock necessary to register new GC block.
* src/rtl/hbsocket.c
* added support MAC addreses array returned by hb_socketGetIFaces()
in BSD based systems like FreeBSD or Darwin (Mac OS X)
* src/vm/hvm.c
% small simplification i FOR EACH initial code
* contrib/hbwin/win_bmp.c
* contrib/hbwin/win_prn2.c
* use Harbour FILE IO API instead of local file access
* contrib/hbct/ctstrfil.c
* formatting
* contrib/gtqtc/gtqtc1.cpp
! fixed possible missing initialization for drawing character images
* include/hbcompdf.h
* include/hbexprb.c
* src/common/funcid.c
* src/vm/hvm.c
% optimize Array() function by replacing function call with HB_P_ARRAYDIM
PCODE. Because this optimization changes RTE and interacts with broken
code like
aVal := Array( 10, 0, "A" )
then it's not enabled by default. User can enable it by -ko switch.
* utils/hbtest/rt_array.prg
* use #pragma -ko- to for tests which interacts with above modification
* src/rtl/hbproces.c
% optimize memory allocation for redirected STDOUT and STDERR buffers in
hb_processRun() function.
The previous implementation was extremly inneficient when
hb_processRun() was used to extract very long output from
child process.
* src/rtl/hbtoken.c
+ added support for dividing text into lines using EOLs used by different
platform. To enable it it's enough to specify .T. as delimiter.
* src/rtl/filesys.c
* pacified MSVC warning - in fact this modification is significant only
for bugy code which changes current directory in MT programs. It's bug
on all platforms using current directory as process not thread
attribute (common behavior).
* include/inkey.ch
* src/rtl/hbgtcore.c
+ added new extended keycodes:
HB_K_TERMINATE
HB_K_MENU
* contrib/gtqtc/gtqtc1.cpp
+ center and rescale to console window dimension keeping aspect size ratio
picture passed to HB_GTI_DISPIMAGE when second parameter is .T.
+ added support for HB_K_MENU key
! fixed dynamic font size modification in fullscreen, maximized and
HB_GTI_RESIZEMODE_ROWS modes
* src/vm/estack.c
* always set NIL in item allocated by hb_stackAllocItem() when it's
used as function. In internal macro this step is missing so allocated
item still may contain other simple value. All places in HVM where
hb_stackAllocItem() is used are safe for such situation (I've just
verified them again).
It fixes problem reported by Mindaugas when hb_stackAllocItem()
was used inside DBFFPT as temporary item and could unintentionally
change memo type if allocated item containes previous numeric
value in valid memo type range.
* src/rdd/wafunc.c
* src/vm/hvm.c
* src/vm/memvars.c
* clear passed item before RTE when accessed memvar or field does not
exists
* src/vm/hashfunc.c
! set default value before decoding 2-nd logical parameter in
hb_HAutoAdd() function. It fixes ambiguity reported by Mindaugas.
* include/hbvm.h
* src/vm/hvm.c
+ added new HVM function: hb_vmTryEval()
This function tries to execute given block or function
inside BEGIN SEQUENCE / RECOVER oErr / END SEQUENCE statement.
It return HB_TRUE if it was executed without any problem and
HB_FALSE otherwise. Result or RTE object is stored in 1-st
parameter and should be freed by the caller using hb_itemRelese()
* src/debug/dbgentry.c
! use hb_vmTryEval() instead of hb_itemDo() to evaluate trace point
expressions. Now debugger does not crash when user sets tracepoint
expression which cannot be evaluated in executed context.
! allow to use expressions with more then 99 variables
! added missing protections against GPF when wrong parameters are
passed to __dbg*() functions
* keep original trace and watch point expression for farther
processing
* replaced IS_IDENT_*() macros with HB_IS*IDCHAR()
+ added new debuger function:
__dbgGetWatchPoints() -> <aWatchTracePoints>
<aWatchTracePoints> contains subarrays describing trace and watch
points:
{ <cExpression>, <lTracePoint>, <lValid>, <xResult> }
It was the last function necessary to extract all information
about debugged context from core internals. Now it's possible
to create fully functional standalone debugger without storing
and updating locally any information about debugger state.
All such information can be accessed at any time using __dbg*()
functions.
It should greatly simplify creating 3-rd party debuggers like
HWDebug or HBQTDebug.
* src/debug/debugger.prg
! fixed callstack window updating during tracing and stepping.
Tanks to Rolf for reporting the problem.
* contrib/hbtip/mail.prg
! fixed missing 1-st adress in CC and BCC parameters - many
for information about the bug and patch
* contrib/hbtip/sendmail.prg
! typo in comment - thanks to Heinz V Bergen too
* src/rtl/gtwvt/gtwvt.c
* some minor modification to sync code with Vitkor's branch
* src/vm/hvm.c
* src/vm/itemapi.c
* small optimization in string comparison
* src/rtl/filebuf.c
! unlock HVM stack before locking local mutex and calling hb_fs*()
functions which also unlocks HVM. It fixes possible deadlock condition
when hb_gcAll( .T. ) is executed.
* include/hbapi.h
* src/vm/garbage.c
! changed mark function semantic.
Adding support for user defined mark function I created race condition
in MT GC code. It happens because blocks marked as deleted were not
scanned by GC mark code so their subitems where not accessible.
To fix it we have to change mark function semantic. Now mark function
can be executed also for blocks currently deleted. It means that GC
block destructor should clean references to just removed items and
subblocks. The best place to make it is clearing pointers to GC blocks
just after hb_itemRelease() or hb_gcRefFree().
It is save to clean the reference just before hb_itemRelease() or
hb_gcRefFree() but only for the single block passed to these functions.
It is not save to clear reference to more then one block and then
execute above functions.
+ check reference count after destructor execution for all blocks
not only arrays - warning it may exploit some wrong C code.
- removed not longer used hb_gcRefCheck() function.
* include/hbvm.h
* src/vm/hvm.c
+ added new internal function hb_vmLockForce()
! fixed to mark GC blocks with active threads
* src/vm/thread.c
* include/hbthread.h
% modified sync mutexes used by xBase++ signal class emulation
to not use item array internally
* updated mutex destructor to new GC mark semantic
! mark GT items if GT is bound with thread item
! fixed mutex notify/subscribe code to not change GC items
when HVM stack is unlocked
! use hb_vmLockForce() to eliminate potential deadlock
+ added new C function hb_threadEnterCriticalSectionGC().
It should be used instead of hb_threadEnterCriticalSection()
in code which manipulates GC items inside critical section.
It's slower but eliminates possible deadlock condition.
Please remember that remember that both functions cannot
be used for the same mutex. So if it's necessary to use
hb_threadEnterCriticalSectionGC() in one place then it
has to be used in all others when the same mutex is locked.
* src/vm/arrays.c
* src/vm/codebloc.c
* src/vm/hashes.c
* src/rtl/hbgtcore.c
* contrib/hbcurl/core.c
* contrib/hbexpat/core.c
* updated destructors to new GC mark semantic
* contrib/hbxpp/dllx.c
! fixed destructor for pointer item created by DllPrepareCall()
Now it respects reference counter.
; added note about real behavior of library handle destructor,
I haven't changed existing code behavior but maybe it should
be done.
* contrib/xhb/hbserv.c
* src/debug/dbgentry.c
* src/rdd/wacore.c
! fixed possible deadlocks by using hb_threadEnterCriticalSectionGC()
instead of hb_threadEnterCriticalSection().
* contrib/hbmemio/memio.c
! slightly modified hb_memfsDirectory() code to avoid race condition
without using hb_threadEnterCriticalSectionGC()
* src/vm/fm.c
% reduce the lock range in HB_FM_STAT builds
* src/rtl/hbsocket.c
! added missing HVM stack unlocking in hb_socketSelect() function
* src/rtl/gtxwc/gtxwc.c
! added few missing locks for version compiled with HB_XWC_XLIB_NEEDLOCKS
* src/rdd/dbtotal.prg
* allow to use symbols instead of codeblocks
* modified Harbour extension which uses ordKey() as default group
signature to work also without index. In such case all records
are summarized into single one.
; Most of above modifications were critical for stability of MT programs.
They should allow to activate GC in any place.
* src/vm/garbage.c
! added protection for activating GC from thread with unlocked HVM stack
when other thread is already executing GC passes. Such situation does
not happen in core code and the protections is only just in case.
* src/vm/hvm.c
! do not release HVM thread STOP signal when GC mark pass is activated.
It was very serious bug which could cause memory corruption when
thread unlocked HVM stack what allow to start GC mark pass by other
thread and then locked HVM and begin to change complex variables
still scanned by the GC thread. This problem have existed from the
beginning when I committed initial version of Harbour MT support.
I do not remember why I added code to remove the STOP signal - probably
I was experimenting with some new things working on Harbour MT code
and I forgot to deactivate it.
NOTE: 3-rd party developers who use GC collectible pointer items
must not activate any PRG code execution inside their mark
functions.
* src/vm/hvm.c
* clear STATIC variables before releasing thread specific data.
It should resolve potential problems with code using TSD to activate
cleanup procedure and later cannot correctly execute pointer item
destructors, i.e. OLE subsystem.
* include/hbvm.h
* src/vm/hvm.c
* added new 7-th parameter fOverLoad to hb_vmRegisterSymbols() function.
When it's TRUE then new public function symbols overload existing ones.
* src/vm/runner.c
! use new fOverLoad parameter of hb_vmRegisterSymbols() function to
overload existing functions (HB_HRB_BIND_OVERLOAD) instead of making
local modifications in global symbol table. It fixes memory corruption
when the same HRB module is reloaded with HB_HRB_BIND_OVERLOAD parameter.
This was very old bug which has existed from the beginning when I added
support for HB_HRB_BIND_* flags.
* 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/vm/hashes.c
% changed sorting algorithm for hash arrays which keeps item order.
% do not resort hash arrays when KEEPORDER flag is cleared but use
existing internal index to create new order.
* changed the behavior of hb_hSort() function. Now for hash arrays
which keep item order it changes internal item pair position sorting
them. The sort order depends on binary and case sensitivity hash
array flags.
* src/vm/hvm.c
! for literal hash array with repeated keys store the last item
with the given key instead of the first one. It restores
previous behavior.
* contrib/rddads/rddads.hbx
+ added AdsSetIndexDirection()
* contrib/rddads/adsfunc.c
! fixed formatting of AdsSetIndexDirection()
* include/hbapi.h
* include/hbapigt.h
* include/hbvm.h
* src/rtl/inkeyapi.c
* src/vm/cmdarg.c
* src/vm/hvm.c
* replaced hb_cmdargProcessVM() with hb_cmdargProcess() and removed
unused for long time custom user cancel key functionality.
- removed dummy hb_inkeySetCancelKeys() C function.
- removed not used for long time HVM C functions and macros:
hb_vmFlagEnabled(), hb_vmFlagSet(), hb_vmFlagClear(),
HB_VMFLAG_HARBOUR, HB_VMFLAG_ARRSTR
* include/hbgtinfo.ch
+ added new HB_GTI_* actions:
HB_GTI_WINHANDLE - Get console window low level handle
HB_GTI_MOUSEPOS_XY - Get mouse position in pixels
HB_GTI_DISPIMAGE - Display image with given name
* include/hbapigt.h
* src/rtl/inkeyapi.c
+ added new C function: int hb_inkeyKeyMod( int iKey )
+ added code for translation from extended key codes to
standard Clipper ones
* include/harbour.hbx
* src/rtl/inkey.c
+ added new PRG function hb_keyMod( <nExtKey> ) -> <nModifiers>
It extract keyboard modifiers (HB_KF_*) from extended key code.
* include/hbgtcore.h
* src/rtl/hbgtcore.c
+ added new macros for generating extended key codes with keyboard
modifiers for ASCII and UNICODE characters
% strip repeated extended key codes for mouse move
! use standard key codes in extended ALERT() GT method
* include/inkey.ch
+ added key modifiers/flags for extended key codes: HB_KF_*
+ added extended key codes: HB_KX_*
- removed unused HB_K_MULTICODE macro
* src/rtl/gttrm/gttrm.c
* replaced old keyboard and mouse code with new one using
extended key codes for keyboard and mouse
+ added support for keyboard modifiers in GPM events
+ added support for keyboard modifiers in terminal mouse events
* src/rtl/gtwvt/gtwvt.h
* src/rtl/gtwvt/gtwvt.c
! invert selection region in WinCE builds
! fixed selection in fullscreen or maximized modes with margins
* do not add EOL to one line selections
+ added support for painting box drawing characters.
This mode can be enabled by:
hb_gtInfo( HB_GTI_FONTATTRIBUTE, HB_GTI_FONTA_DRAWBOX )
and it causes that GTWVT draws all box characters as pixmap
regardles of used font.
So far this functionality existed only in GTXWC.
+ added support for HB_GTI_FONTA_FIXMETRIC and HB_GTI_FONTA_CLRBKG
So far this functionality existed only in GTXWC.
+ added support for HB_GTI_WINHANDLE
! do not add EOL at the end of one line selection
* src/rtl/gtxwc/gtxwc.h
* src/rtl/gtxwc/gtxwc.c
* replaced old keyboard and mouse code with new one using
extended key codes for keyboard and mouse
* renamed HB_GTXWC_* box drawing macros to HB_BOXCH_* ones
* do not add EOL to one line selections
! fixed few keypad keycodes
* use cursor blinking rate HB_GTI_CURSORBLINKRATE for whole
show/hide period not only half of them. If someone use this
GTI action in his code then now he should double passed
value. [incompatible]
! fixed return value for HB_GTI_SETPOS_ROWCOL
* few minor cleanups
* tests/gtkeys.prg
* replaced old keyboard and mouse code with new one using
extended key codes for keyboard and mouse
+ added description for some key codes
+ added some initial HB_GTI_* actions
+ change cursor size by K_INS
+ display terminal size after HB_K_RESIZE
+ display mouse position after mouse key codes
! display all characters created by hb_keyChar()
! fixed formatting
* src/rdd/dbf1.c
* added more general protection against GPF in not fully
initialized workarea
* (all files)
* stripped svn header
* minor cleanups
; use following command to find out the history of files:
git log
git log --follow
git blame
git annotate
