* harbour/include/hbstack.h
% enable assembler inline function to access HVM thread stack pointer
without native compiler TLS support even if HB_STACK_PRELOAD is not
defined in MinGW builds. It gives some additional small speed
improvement in MT mode
* harbour/source/rtl/idle.c
* unlock HVM inside hb_releaseCPU() function
* harbour/contrib/hbwin/win_ole.c
! fixed casting and C++ compilation in some compilers
* harbour/include/Makefile
+ harbour/contrib/examples/rdddbt/hbrdddbt.h
- harbour/include/hbrdddbt.h
! moved hbrdddbt.h to correct location
* harbour/include/Makefile
! added missinf hbthread.h
* harbour/tests/rddtest/rddmktst.prg
* harbour/tests/rddtest/adscl52.prg
* harbour/tests/rddtest/adscl53.prg
* harbour/tests/rddtest/ntxcl52.prg
* harbour/tests/rddtest/ntxcl53.prg
* harbour/tests/rddtest/cdxcl52.prg
* harbour/tests/rddtest/rddtst.prg
* harbour/tests/rddtest/cdxcl53.prg
! fixed description in header I wrongly copied without updating
from other files
+ harbour/tests/hsxtest.prg
+ added test code for HiPer-SEEK indexes
* harbour/source/rtl/filesys.c
+ added new functions: hb_fsExtName(), hb_fsReadAt(), hb_fsWriteAt(),
hb_fsTruncAt() to use with shared file handles.
Please test it in other *nixes. I used pread[64]()/pwrite[64]()
POSIX functions but I do not know if they are available by default
in other supported platforms like MacOSX, BSD, HPUX, SunOS.
If not they I will have to implement some workaround.
* harbour/common.mak
* harbour/source/rtl/Makefile
* harbour/include/hbapifs.h
+ harbour/source/rtl/filebuf.c
+ added set of new file functions (hb_file*()) which are designed to
use with shared file handles and locks (buffers in the future):
hb_fileExtOpen(), hb_fileCreateTemp(), hb_fileClose(), hb_fileLock(),
hb_fileReadAt(), hb_fileWriteAt(), hb_fileTruncAt(), hb_fileSize(),
hb_fileCommit(), hb_fileHandle()
These functions operate on PHB_FILE structure instead of HB_FHANDLE.
Now in POSIX systems they share file handle between aliased or
simultaneously open by other thread work areas. It resolves the
problem with releasing all process FCNTL locks by any close()
operation even on duplicated or open() separately handle.
Now DOS deny flags emulation works in single process even if
BSD locks are disabled (non Linux *nixes).
They also keep internal file lock list what allows to synchronize
threads and aliases with file locks in POSIX systems. In the future
it will be used also for internal locking to synchronize threads
without OS level locks - pseudo exclusive mode in cloned WA.
Finally these structures will be used also for file buffers
which will be shared between aliased WA and threads.
* harbour/include/hbrdddbf.h
* harbour/include/hbrddcdx.h
* harbour/include/hbrddntx.h
* harbour/include/hbrdddel.h
* harbour/include/hbrddsdf.h
* harbour/source/rdd/dbf1.c
* harbour/source/rdd/delim1.c
* harbour/source/rdd/sdf1.c
* harbour/source/rdd/dbffpt/dbffpt1.c
* harbour/source/rdd/dbfntx/dbfntx1.c
* harbour/source/rdd/dbfcdx/dbfcdx1.c
* harbour/source/rdd/hsx/hsx.c
* harbour/contrib/hbbmcdx/bmdbfcdx.c
* harbour/contrib/hbbmcdx/hbbmcdx.h
* use PHB_FILE and hb_file*() functions instead of HB_FHANDLE (hb_fs*())
to access files.
+ harbour/tests/aliaslck.prg
+ added test code for file lock synchronization between aliased
work areas
* harbour/include/hbver.h
* changed version number and ChangeLog entry
* harbour/make_tgz.sh
! fixed creating PP in DJGPP builds
* harbour/contrib/dot/pp.prg
! minor fix in parameter checking
* harbour/contrib/libmisc/hb_f.c
! fixed bug in record numbering after reaching EOF position and
backward skipping
* harbour/include/hbrdddel.h
* harbour/source/rdd/delim1.c
+ implemented DELIM RDD. It should be fully Clipper compatible
in comma files parsing. If you will find any difference, please
inform me and I'll update it. Unlike in Clipper in [x]Harbour
these RDD can be used also in direct import operations with
APPEND FROM ... VIA "DELIM"
These RDD also support whole record transfering between RDDs
if they support the same record format and field representation
and can translate extended field types like "I", "B", "V3", "V4",
etc.
* harbour/source/rtl/dbdelim.prg
* replaced old code of __dbDelim() function with simple calls to
__dbCopy()/__dbApp() - it should fix most of the reported
incompatibilities in DELIM import/export operations and
increase their speed many times
* harbour/include/hbrddsdf.h
* harbour/source/rdd/sdf1.c
+ implemented SDF RDD. It should be fully Clipper compatible
in SDF files parsing. If you will find any difference, please
inform me and I'll update it. Unlike in Clipper in [x]Harbour
these RDD can be used also in direct import operations with
APPEND FROM ... VIA "SDF"
These RDD also support whole record transfering between RDDs
if they support the same record format and field representation
and can translate extended field types like "I", "B", "V3", "V4",
etc.
* harbour/source/rtl/dbsdf.prg
* replaced old code of __dbSDF() function with simple calls to
__dbCopy()/__dbApp() - it should fix most of the reported
incompatibilities in SDF import/export operations and
increase their speed many times
* harbour/source/rdd/dbcmd.c
+ added support for table structure inheriting in transfer RDDs.
+ added eleventh parameter with delimiter to __dbCopy()/__dbApp()
* harbour/source/rdd/dbf1.c
* force 10 bytes size for memo fields in DBF and SMT memos.
* some minor clean-up.
* harbour/source/rdd/workarea.c
* added some dummy DBI_* actions in base RDD level
