* harbour/include/hbthread.h
* added workaround for different CRITICAL_SECTION definitions in
some compiler header files
* harbour/source/vm/fm.c
* harbour/source/hbzlib/ChangeLog
* harbour/source/hbzlib/zutil.c
* use #pragma warning to pacify OpenWatcom warnings
* include/hbdefs.h
! Further tweaked HB_STRICT_ALIGNMENT detection by using
HB_CPU_* macros.
Please review and extend/tweak if needed.
* include/hbdefs.h
* include/hbsetup.h
* include/hbinit.h
* source/vm/fm.c
* source/common/hbver.c
* config/sunos/sunpro.cf
+ Patch for Sun support by Tamas Tevesz.
- SunPro compiler C++ mode support.
- HB_STRICT_ALIGNMENT config refinement for Sun platform.
- SunPro compiler version detection will detect raw version
number for newer releases.
- SunPro optimization flag tweaks for C/C++ mode.
Trying to be Blastwave.org compatible as much as possible.
- LD_OUT -o option tweak for SunPro.
; NOTE: -o option should better be revised for all supported
[gcc] compilers to have space between it and the
filename.
* source/vm/extrap.c
+ Changed to use HB_CPU_* macros instead of using compiler
specific ones.
* INSTALL
* bin/postinst.sh
* source/vm/fm.c
* source/common/hbver.c
+ config/sunos/sunpro64.cf
* config/sunos/sunpro.cf
* ChangeLog
+ Applied second SunPro patch sent by Tamas Tevesz.
Addressed these former TODOs:
+ Added sunpro64 target.
+ Finalized C compiler options (including warning level settings).
+ Added compiler identification code to hbver.c.
+ dynamic lib creation.
* utils/hbmk2/hbmk2.prg
! Fixed RTE when run on sunos. Needs further work to discover how
it could run onto that codepath, normally it shouldn't.
* bin/postinst.sh
% Cleaned a few remaining 'gpp' HB_COMPILER value checks.
* include/hbdefs.h
* include/hbsetup.h
* include/hbinit.h
* source/vm/fm.c
+ config/sunos/sunpro.cf
+ Added first round of Sun Pro (Sun Studio) compiler.
Contribution of Tamas Tevesz.
; TODO:
+ Add other Sun Pro CPU targets.
* Finalize __inline keyword usage. (currently in fm.c)
Przemek, could you take a look at it?
! Fix warning option.
+ Add compiler identification code to hbver.c.
+ hbmk2 support.
+ dynamic lib creation.
* utils/hbmk2/hbmk2.prg
* Minor rename.
* contrib/hbclipsm/gauge.c
% Minor opt, plus removed cast.
* contrib/hbwin/wce_smsc.c
* contrib/hbwin/wce_sim.prg
* Minor formatting.
* source/vm/fm.c
! Fixed bcc C++ build errors by enabling USE_DL_PREFIX in this case.
Only this warning remain:
Warning W8008 ../../fm.c 386: Condition is always true in function dlmalloc_destroy()
+ harbour/include/hbtask.h
+ harbour/source/vm/task.c
* harbour/include/hbthread.h
* harbour/include/hbatomic.h
* harbour/source/vm/thread.c
* harbour/source/vm/hvm.c
* harbour/source/vm/fm.c
* harbour/source/rtl/idle.c
* harbour/source/rtl/filesys.c
+ implemented OS independent task switching system - it gives PTHREAD
compatible basic API so it can be used in HVM as alternative MT support
which does not use any OS threads. As long as Harbour does not call
any blocking OS function then it's possible to create and execute
simultaneously many threads though only one CPU is used and switched
between HVM threads. It gives similar scalability to xbase++ threads
and also similar behavior in item protection at .prg level.
Now it's possible to use HVM threads in any OS.
Of course it does not mean that Harbour adds in some magic way
thread support to OS-es which does not support threads like DOS.
It only means that HVM supports threads for .prg code just like
in native MT environment as long as some C code does not block
task switching or process execution will not be frozen by sth, i.e.
executing other process (__run()) in single process OS like DOS.
In some cases it can be interesting alternative even in OS which
have native thread support.
All tests/mttest*.prg programs and speedtst --thread=<n> --scale
are executed correctly with new task switching just like with
OS native MT support.
Compilation with task switching in hbvmmt library can be forced
by HB_TASK_THREAD macro which also disable native OS threads
support.
For task context switching two alternative methods are used:
1) getcontext()/makecontext()/swapcontext() (SUSv2, POSIX.1-2001)
which is preferable because does not need any additional
hacks but not all OS-es supports these functions.
It's enabled by default in Linux builds.
2) setjmp()/longjmp() (POSIX, ISO 9899 (C99)) otherwise.
These functions are supported by most of C compilers
but there is no function to set new stack in saved context
so it's necessary to introduce for each architecure/C compiler
peace of code which makes it. Macro HB_TASK_STACK_INIT() in
task.c makes it. I defined this macro for x86@32 in DJGPP
Linux GCC and OpenWatcom builds. I tested OpenWatcom builds only
in DOS and Linux but probably it works in all x86@32 builds.
If someone is interesting in adding support for some other
platforms which does not support ucontext.h and 1-st methods
then please define above macro for them.
Have a fun with new toy ;-)
* harbour/source/vm/Makefile
* enabled hbvmmt in DJGPP and OpenWatcom DOS builds. It works well.
Viktor if possible please add support for -mt switch in hbmk2
in all builds even if we do not compile hbvmmt by default so
it can be used with DJGPP and OW and any other builds for which
someone enable hbtask.c though OS does not support threads.
* harbour/contrib/hbmzip/hbmzip.c
! fixed '[const] char|BYTE *' casting in DOS and OS2 builds
* include/hbcomp.h
* include/hbcompdf.h
* include/hbapifs.h
* include/hbexprop.h
* include/hbmacro.h
* include/hbapi.h
* include/hbapiitm.h
* include/hbdate.h
* source/pp/ppcore.c
* source/vm/itemapi.c
* source/vm/arrays.c
* source/vm/extend.c
* source/common/expropt1.c
* source/rtl/philes.c
* source/rtl/empty.c
* source/rtl/minmax.c
* source/rtl/dates.c
* source/rtl/filesys.c
* source/rdd/dbfnsx/dbfnsx1.c
* source/rdd/dbfcdx/dbfcdx1.c
* source/rdd/dbf1.c
* source/rdd/dbffpt/dbffpt1.c
* source/rdd/hbsix/sxdate.c
* source/compiler/hbmain.c
* source/compiler/complex.c
* source/compiler/harbour.yyc
* source/compiler/harbour.y
* contrib/hbct/files.c
* contrib/hbct/dattime2.c
* contrib/hbct/datetime.c
* contrib/xhb/filestat.c
* contrib/hbtip/utils.c
* examples/hboleold/win_ole.c
* LONG -> long for date/time related values
(julian, date, type, millisec)
; This is an effort to clean Harbour sources from
Windows/legacy-Clipper basic types and replace
them with own or ANSI C ones.
In the above case, usage wasn't consistent,
LONG and long were mixed.
* source/main/harbour.c
* source/vm/fm.c
* LONG -> long for some stat counters.
* INSTALL
+ Added iccia64 build example.
* source/vm/fm.c
+ Added C++ new/delete operator override to hb_xgrab()/hb_xfree().
Thanks to Xavi.
Please make any required correction, I didn't do tests with it.
* utils/hbmk2/hbmk2.prg
+ Added support for linux/icc compiler. (untested)
; Tested win/icc support (static exe, shared exe, lib, dll).
It worked, except that in shared mode it will display this,
when the MinGW built .dll is used (instead of its own):
Unrecoverable error 9012: Can't locate the starting procedure: 'MAIN'
* source/vm/fm.c
! Renamed s_fInited to s_fInitedFM to avoid collision
with similarly named variable in estack.c when using
HB_FM_STATISTICS and HB_HVM_ALL at the same time.
Thanks Xavi for the find.
* source/vm/Makefile
+ Enabled "amalgamated" VM lib for all compilers/platforms,
except for those where this doesn't give any advantage, or
it breaks the build, these are: pocc*, xcc, bcc, dmc.
Please note that Open Watcom 1.7 might also fail (I didn't
test), so make sure to upgrade to 1.8.
* harbour-win-spec
* harbour-wce-spec
* mpkg_win.bat
* mpkg_tgz.sh
* harbour.spec
* bin/hb-func.sh
* include/hbsetup.h
* source/vm/Makefile
* source/vm/fm.c
- source/vm/fm
- source/vm/fmmt
* utils/hbmk2/hbmk2.prg
- Removed hbfm and hbfmmt lib references. These libs are
no longer generated or used.
* Narrowed HB_FM_STATISTICS presence in various source
files. From now on the system default for all builds is
HB_FM_STATISTICS_OFF. We may change this to
HB_FM_STATISTICS_DYN_OFF if we find a solution to the
potential and actual problems with this method.
If someone needs to change mem stat default, it can be
done by ususal means.
* harbour/source/vm/hvm.c
* harbour/source/vm/fm.c
* harbour/source/vm/cmdarg.c
! fixed GPF when hb_vmInit() is called without initializing
application startup parameters by hb_cmdargInit().
Thanks to Enrico's friends for information about the problem.
* some code cleanup
* harbour/source/vm/dlmalloc.c
! fixed casting
* harbour/source/vm/fm.c
% use DLMALLOC as default memory manager in OpenWatcom Windows builds.
Warning: HB_FM_DLMT_ALLOC does not work with OpenWatcom.
* harbour/config/dos/owatcom.cf
* harbour/config/win/owatcom.cf
* harbour/config/linux/owatcom.cf
* harbour/config/os2/owatcom.cf
* changed alignment settings from 8 to 4
* synced optimization flags used in different builds
! fixed linker parameters when HB_BUILD_DEBUG=yes is set
; TOVERIFY: I do not remember what calling convention should be used
in OpenWatcom MT OS2 builds to eliminate GPF when APIENTRY16 functions
are called. It's possible that current settings are wrong.
* harbour/config/win/owatcom.cf
* removed STACK=65536 from OpenWatcom linker parameters.
This switch probably was inherited from old DOS WatcomC builds.
Later I'll test current OpwnWatcom builds to check if we can also
remove it from DOS builds when cwstub.exe is used.
* restored -bm switch. When DLMALLOC is used it does not cause
noticeable performance reduction.
* use default register calling convention. We can change it
in the future but 1-st we should check the performance overhead.
If calling convention is a problem only for .DLLs then it can
be resolved by modifying HB_EXTERN declaration and adding function
attributes which will force given calling convention.
* harbour/config/linux/owatcom.cf
* enabled pentium pro instruction in OpenWatcom Linux builds.
It reduces the code size and increase performance a little bit.
Windows users which do not need pure pentium CPU support can make
the same.
* harbour/utils/hbmk2/hbmk2.prg
! fixed object extension used in OpenWatcom Linux builds: it's .o not .obj
! fixed linker parameters in OpenWatcom Linux builds: DEBUG ALL is single
option
! fixed OpenWatcom calling convention settings. Viktor you cannot chose
simultaneously register and stack calling convention. You have to chose
one and keep it synced with Harbour compile time settings. Otherwise
you will have unresolved external or you will force creating by linker
dynamically function call wrappers (of course if OW support such
functionality) what may strongly reduce the performance.
! removed -j compile time switch - it's not used to compile core code
* synced optimization flags with core code
I'm really interested in current OpenWatcom speedtst results compared
with other Windows builds (MSVC, MinGW, BCC, POCC) in ST and MT builds.
* harbour/include/hbapi.h
* harbour/include/hbstack.h
* harbour/source/vm/estack.c
* harbour/source/vm/fm.c
+ added support for thread local memory allocators by using MSPACE
mechanism in DLALLOC.
It can be enabled by compiling Harbour with HB_FM_DLMT_ALLOC macro.
By default up to 16 (HB_MSPACE_COUNT macro) independent memory
allocators is dynamically created for new threads. When more threads
is used then allocators are shared between them and HVM tries to
balance used allocators to reduce number of threads using the
same allocator.
It should give noticeable scalability improvement in some OS-es
where default memory manages has pure MT performance.
Please try to compiler speedtst.prg with Harbour compiled with
and without HB_FM_DLMT_ALLOC and compare results in this test:
speedtst --thread=2 --scale
using real multiCPU machine.
* harbour/source/vm/dlmalloc.c
* pacified warning when FOOTERS are enabled
* harbour/source/vm/garbage.c
* cleanup comment
* source/vm/fm.c
! Fixed compilation error in dlmalloc with pocc64.
* config/win/pocc64.cf
* config/win/msvc64.cf
* Cleanup.
* config/win/gcc.cf
! Added gdi32 to system lib list.
* config/win/mingw.cf
* config/win/mingwce.cf
* Minor sync with each other.
% Two duplicated lines removed from mingwce. I hope these
weren't intentional.
---
AR = $(CCPREFIX)ar
ARFLAGS = $(HB_USER_AFLAGS)
---
* harbour/include/hbsetup.h
* minor modifications in macros
* disabled some GCC function attributes when ICC is used
* harbour/source/pp/ppcore.c
* casting
* harbour/source/vm/fm.c
* destroy dlmalloc mspace in hb_xexit() when USE_DL_PREFIX is used
otherwise call malloc_trim(0) to at least reduce allocated memory
* source/rtl/gtwvt/gtwvt.c
* BYTE -> CHAR (Windows) - UNDONE.
* include/hbdefs.h
+ Added (commented) new types. Work in progress, incomplete.
* contrib/hbssl/ssl.c
- Temporary disabled SSL_GET_READ_AHEAD() as a workaround for
BCC linking problem with SSL_get_read_ahead() OpenSSL
library function (symbol is unresolved due to missing
leading underscore in implib generated ssleay32.lib file).
* source/vm/set.c
% hb_fsFile() -> hb_fsFileExists()
* source/rtl/hbffind.c
* HB_WIN_32_MATCH -> HB_WIN_MATCH
* doc/whatsnew.txt
* include/hbsetup.h
* source/vm/fm.c
* #define HB_FM_WIN32_ALLOC renamed to HB_FM_WIN_ALLOC.
; INCOMPATIBLE: For those who are using this switch, please
update your make files.
* source/rtl/tpopup.prg
! Fixed to draw separator lines with BOX GT attributes,
so that they appear properly.
* harbour/source/vm/fm.c
! fixed typo in default FM module setting/
* harbour/include/hbapicls.h
* harbour/source/vm/classes.c
* harbour/source/debug/dbgentry.c
* renamed hb_dbgObjSendMessage() to hb_dbg_objSendMessage() to keep
naming convention used for other C debugger functions
* harbour/source/vm/debug.c
* added information about existing debugger functions and where they
are used.
* harbour/include/hbvm.h
* harbour/include/hbapidbg.h
* harbour/include/hbvmpub.h
* harbour/include/hbstack.h
* harbour/include/hbapi.h
* harbour/source/vm/codebloc.c
* harbour/source/vm/estack.c
* harbour/source/vm/hvm.c
* harbour/source/debug/dbgentry.c
* changed the method of static variable handling. Now we do not
have centralized array of all statics used by application (s_aStatics)
but each module uses it's own array to keep statics. The old
s_aStatics has been removed. It means that loading new PCODE module
from .hrb file or shared/dynamic library does not cause any static
table reallocation so it does not creates potential GPF trap for MT
code which may operate on static vars.
This modification changed a little bit the debugger interface.
I also updated some unused by us debug functions like __DBGVMVARSLIST()
or __DBGVMVARSLEN() anyhow any 3-rd party debuggers should be
updated for new API and modified hb_dbgEntry() syntax.
* 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/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.
* source/rtl/gtwvt/gtwvt.c
! Fix to previous fix to maximize the window when the initial
size would be too big to fit on screen (stupid omission).
Now the previously cited sizing problem is alright.
* source/vm/fm.c
+ Replaced HB_NO_FM_STAT and HB_FM_STAT envvars
with HB_FM_STAT=yes|no (maybe 0|1 or on|off woud be
better or could be accepted as alternatives or
replacement). This is to avoid ambiguity when both
envvars were set or the app default is not known.
+ Added HB_FM_STATISTICS_DYN_OFF to turn off mem tracking
by default when stat module is linked in.
* source/common/hbverdsp.c
* Minor cosmetic.
* harbour/source/vm/fm.c
+ added support for enabling/disabling memory statistic when
application starts. To enable memory statistic it's enough
to set HB_FM_STAT environment variable to non empty value.
To disable memory statistic it's enough to set HB_NO_FM_STAT
envvar. The default setting may depend on build time option.
Now is enabled when memory statistic is compiled when harbour
is created. It may not work if C compiler does not allow to
check environment variables from startup code. F.e. functions
to access environment can be also part of dynamic library
registered after harbour.dll or harbour core code startup
functions.
Viktor I would like to ask you only about one thing.
Please do not remove current support for HB_FM_STATISTICS_OFF
from GNU make builds so Harbour user can still create binaries
without any memory statistic code using -[no]fmstat switch in
hb* scripts. The above feature is only for users who need support
for enabling/disabling memory statistic to test application
if foreign environment, f.e. at client computers. Of course if
it will work with given C compiler and/or type of application
(linked dynamically or statically). Please check.
* source/vm/fm.c
! Fixed missing abort() on WinCE. (for Pelles C only, but
the fix should work for any WinCE compilers).
; TOFIX: Now the only remaining showstopper for PellesC5 WinCE
support is:
../../math.c(652): error #3120: [asm] Symbol 'inf' is undefined.
* harbour/include/hbapi.h
* harbour/source/vm/hashes.c
* harbour/source/vm/hashfunc.c
* changed hb_hSort() to execute sorting when hash is marked as needed
resorting instead of marking hash for resorting
* harbour/source/vm/fm.c
* formatting
* harbour/contrib/hbdbgfx/dbgfxc.c
! fixed casting in ASCII build
+ added support for UNICODE builds
* harbour/contrib/gtwvg/gtwvg.c
* cover LWA_ALPHA usage by #if ( _WIN32_WINNT >= 0x0500 )
to keep at least compile time compatibility with older systems
* common.mak
* source/vm/Makefile
* source/vm/vmmt/Makefile
* source/vm/fm.c
+ source/vm/fmhb.c
+ source/vm/hbmem.c
+ Moved functions not depending on FM_STATISTICS setting,
to separate files.
; NOTE: MEMORY() function should IMO move to rtl, and
hb_xmem*() functions to common lib.
* harbour/source/vm/garbage.c
* unblock HVM before executing destructors and releasing blocks
* harbour/source/vm/fm.c
! do not call internal error with active lock when FM statistic
is enabled
* harbour/source/vm/thread.c
! stop other threads in hb_threadOnce() if the 1-st one
executes bAction block to be sure that it will be completed
before thread leave hb_threadOnce() function
* harbour/source/vm/set.c
! do not call hb_inkeyReset() in set release. It's not necessary
and it could change keyboard buffer size for other threads using
the same GT driver.
* harbour/source/vm/hvm.c
% allocate new thread stack without lock
* harbour/source/rtl/hbgtcore.c
* harbour/source/rtl/gtapi.c
* moved clipboard resetting from hbgtcore.c to gtapi.c
* harbour/source/rtl/filesys.c
! do not use read/write operations with given file offset in Win9X
which seems to not support such functionality
* harbour/include/Makefile
+ harbour/include/hbatomic.h
* harbour/include/hbthread.h
* harbour/source/vm/garbage.c
* harbour/source/vm/fm.c
* moved atomic and spinlock functions into hbatomic.h
* harbour/include/hbatomic.h
+ added atomic inc/dec for GCC and x86@64 and PPC@32
+ use OSAtomic*() and OSSpin*() functions for atomic inc/dec and
spinlocks in Darwin builds
+ added spinlocks to MS-Win builds
* harbour/source/vm/fm.c
! do not build DL-MM when HB_FM_WIN32_ALLOC is set
* harbour/tests/speedtst.prg
+ added support for --exclude=<tests,...> option to exclude some tests,
f.e.:
./speetest --exclude=029.030.023.025.027.040.041.043.052.053.019.022.031.032
disables all tests which allocates memory. The same tests can be also
excluded by --exclude=mem
* source/vm/fm.c
+ Enabled dlmalloc (HB_FM_DL_ALLOC) for MSVC, BCC and MINGW builds.
; It's possible to disable dlmalloc with HB_FM_STD_ALLOC
for testing purposes.
* contrib/rddads/adsmgmnt.c
! Blind fix to ADSMGGETWORKERTHREADACTIVITY() where the
first element in list wasn't returned.
Please test, I couldn't.
* source/vm/fm.c
! Stupid typo. (I guess I'm completely confused by separated
"#" and preprocessor command :/ Second typo in two lines.)
* source/vm/fm.c
! Enabled USE_DL_PREFIX for _MSC_VER compilers.
This may be refined, but without this setting MSVS 2008
would give an error on exit and GPF.
* harbour/source/vm/fm.c
+ harbour/source/vm/dlmalloc.c
+ added alternative memory manager written by Doug Lea
It can be activated by recompiling Harbour with HB_FM_DL_ALLOC macro.
Please make test on different platforms with different C compilers.
Linux users will not benefit from it because the default GLIBC MM
is derived from a version of code.
NOTE: the results may strongly depend on total memory allocation
and will be different then in simple memory tests.
