* config/lib.mk
+ Added hack to make hbpp hack (in source/pp/Makefile) happy.
Should fix missing 'pow' symbols reported on the list.
* config/detect.mk
! Fixed gpm mouse detection prerequisites.
; TODO: Will need some more tweaking to make it clean, I'll do it
a bit later.
* source/rtl/gtcrs/Makefile
* source/rtl/gttrm/Makefile
* source/rtl/gtsln/Makefile
* Using HB_HAS_GPM instead of old HB_GPM_MOUSE.
* config/bin.mk
* config/dyn.mk
* config/bsd/libs.mk
* config/darwin/libs.mk
* config/hpux/libs.mk
* config/dos/djgpp.mk
* config/win/mingw.mk
* config/linux/libs.mk
* config/os2/gcc.mk
* config/sunos/libs.mk
* Extended the way it's decided whether to include
rtl external lib dependencies or not. Now binary specific
logic was moved to bin.mk, and dynamic lib logic was added
to dyn.mk.
The flag is called HB_LINKING_RTL (not empty means yes).
+ Updated syslib list assembly parts to use HB_HAS_*
autodetection variables instead of trying to find it out
by other means.
* config/global.mk
+ Adding gtcrs, gtsln, gtxwc to std liblist if required
components are available. This means that it's now again
possible to select these as build-time default GTs.
* Updated comment for setting plans.
+ config/detfun.mk
+ config/detect.mk
* config/global.mk
+ Added generic external component detection function.
+ Added central detection for optional external components
used by Harbour core. These are: openssl, gpm, slang, curses, x11
Easy to extend with new ones.
The detection code will run once per make session and
results can be used in all our make files by checking HB_HAS_*
variable. If it's empty we cannot use the component, if
it's not we can. In this case it contains dir where headers
were located. It's possible that it's a list of paths.
Users can control these components by using HB_INC_* variable
the following way: if it's left empty, Harbour make system
will automatically look into default locations, this usually
works on *nix systems. If set to a path (or a list of paths),
this list will be checked. Finally to explicitly disable a
component, user can set the variable to 'no'.
Following legacy control variables are yet understood, but
the will be removed in the near future:
HB_WITHOUT_GTCRS=yes is the same as HB_INC_CURSES=no
HB_WITHOUT_SLANG=yes is the same as HB_INC_SLANG=no
HB_WITHOUT_GTXWC=yes is the same as HB_INC_X11=no
Notice that these settings aren't meant to allow user control
of actual Harbour components (like gtxwc). If we need something
like this, we can do it, but it wasn't the subject of this change.
HB_GPM_MOUSE var is still set for compatibility with internals.
NOTE: I've left verbose output on to see what's happening, this will
tuned after testing.
; TODO: Remove reliance on legacy settings in our own codebase.
; TODO: Start using HB_HAS_* values for dynamic lib syslib list assembly
and in GT Makefiles.
* contrib/hbtip/hbtipssl/Makefile
* contrib/hbtip/Makefile
* contrib/hbssl/Makefile
* HB_HAS_OPENSSL works a little differently now, not empty
means 'yes', empty means 'no'.
* config/bsd/gcc.mk
* config/darwin/gcc.mk
* config/darwin/icc.mk
* config/hpux/gcc.mk
* config/sunos/gcc.mk
* config/sunos/sunpro.mk
+ Added "system" libs to dynamic lib creation commands.
; TODO: There is one step left: To build system lib list
also when dynamic libs are to be created.
This will need some extra detection logic.
* INSTALL
+ Added new section: HOW TO DO A PARTIAL [RE]BUILD
This wouldn't have been possible a week ago, the
situation was so much different and complicated for
various cases.
! Cleaned msys/cygwin/<DIR> references and NOTEs.
* config/dyn.mk
* config/bin.mk
* config/global.mk
+ config/bsd/libs.mk
* config/bsd/gcc.mk
+ config/hpux/libs.mk
* config/hpux/gcc.mk
+ config/darwin/libs.mk
* config/darwin/gcc.mk
* config/darwin/icc.mk
+ config/linux/libs.mk
* config/linux/global.mk
+ config/sunos/libs.mk
* config/sunos/gcc.mk
* config/sunos/sunpro.mk
% Moved "system" library logic to compiler libs.mk files
for all *nix compilers.
* config/dos/djgpp.mk
* config/global.mk
* DJGPP with win-based make messages converted to a warning
and moved next to the other similar warning detecting
another non-ideal combination.
* INSTALL
+ Documented HB_CCPATH, HB_CCPREFIX, HB_CCPOSTFIX config variables.
* source/Makefile
* config/lib.mk
+ Added gtxwc to Harbour dynamic libs on *nixes.
* config/global.mk
! Always set HB_INC_INSTALL on sh shells to make postinst.sh
happy.
* config/linux/global.mk
* config/linux/gcc.mk
* config/linux/icc.mk
* config/linux/sunpro.mk
% Moved 'system' library name and path list forming logic to
platform global.mk.
; TODO: Also for other platforms.
* source/Makefile
! Fixed to use '-' separator between dynlib name and version
on *nixes too, except on darwin, where it's a '.'.
* ChangeLog
* config/global.mk
! Added 'SHELL := /bin/sh' lines after 'sh' HB_SHELL detection.
Under certain circumstances SHELL will point to COMSPEC even
when launched from bash shell (f.e. DJGPP GNU Make launched
from DJGPP sh.exe running on Windows OS or using MSYS sh.exe).
This seems to be a safe and often used solution on all platforms
according to GNU Make docs '5.3.1 Choosing the Shell' section:
http://www.gnu.org/software/make/manual/make.html
Fix me if you find this wrong.
* ChangeLog
! Stripped non-ASCII (accented) chars in recent entries.
See: 2008-11-09 12:26 UTC+0100 Viktor Szakats
* harbour/include/hbapifs.h
* harbour/source/rtl/filebuf.c
+ added hb_fileExists() and hb_fileDelete() functions
+ added support for alternative file IO API used in RDD code which
can be implemented by 3-rd party developers and registered using
hb_fileRegister() function.
* harbour/source/rdd/dbf1.c
* harbour/source/rdd/dbffpt/dbffpt1.c
* harbour/source/rdd/dbfntx/dbfntx1.c
* harbour/source/rdd/dbfnsx/dbfnsx1.c
* harbour/source/rdd/dbfcdx/dbfcdx1.c
* use hb_fileExist() and hb_fileDelete() functions for files accessed
by hb_file*() API
* harbour/include/hbvmpub.h
! fixed default size macro used in HB_ITEM_PUT_LONGRAW()
* harbour/include/hbexprb.c
* enable ("alias")-> optimization only when -ko harbour compiler
option is used - it changes RT error.
* config/global.mk
+ Added SVN revision detection and display in log header.
Will kick in only if svn is installed. For non-live SVN
sandboxes (exported/nightly/other dl) it will display
'exported'. It will also show if there was any local
modifications done by appending an 'M' to revno.
* TOFIX -> NOTE
* source/Makefile
% Minor to recent.
* config/global.mk
+ Added dos/djgpp autodetection for *nix platforms.
* contrib/hbfbird/Makefile
+ Extended autodetection and feedback. Experimenting.
* config/globsh.mk
+ Added web links to DJGPP FAQ instead of binary FAQ package.
* contrib/hbqt/Makefile
* contrib/hbfbird/Makefile
* contrib/gtqtc/Makefile
* source/Makefile
* config/global.mk
* HB_OS_UNIX GNU Make var renamed to HB_PLATFORM_UNIX.
Content is now non-empty or empty (was yes or no).
* config/sunos/sunpro.mk
+ Changes made to config/linux/sunpro.mk applied here also.
* harbour/config/linux/sunpro.mk
! fixed dynamic library creation - @<filename> is not supported by
SunPRO C compiler, because it's ported only to POSIX environments
then we can simply pass file list as shell arguments
* harbour/include/hbrdddbf.h
* harbour/source/rdd/dbf1.c
* harbour/source/rdd/dbffpt/dbffpt1.c
* cleaned memo block size decoding.
* harbour/include/hbvmpub.h
+ added new internal macro HB_ITEM_PUT_LONGRAW()
* harbour/source/vm/hvm.c
* harbour/source/vm/itemapi.c
* use HB_ITEM_PUT_LONGRAW()
* INSTALL
+ Added DJGPP cross-builds to support matrix.
+ Added DJGPP bash.exe to the shell mix.
+ Added OS/2 bash.exe to the shell mix.
; NOTE: djgpp build fails here when using bash too
(I've used a Windows machine to test), see below.
* ChangeLog
* Updating pending item status.
* contrib/hbbtree/hb_btree.c
! Added TOFIX about bad bug. (hitting win64, maybe more)
Reported on the list a while ago.
* utils/hbmk2/hbmk2.prg
- Deleted HB_CCPATH support for *nix based gcc/icc compilers.
To be in sync with GNU Make files.
+ Added HB_CCPATH support for os2 and dos.
+ Added HB_CCPREFIX and HB_CCPOSTFIX support for dos/djgpp to
support cross-builds.
+ Using host dependent executable extension for compiler
tools in dos/djgpp to support cross-builds.
* config/wce/mingwarm.mk
* config/win/mingw.mk
+ Added HB_CCPOSTFIX support. Now in sync with rest of gcc
family compilers in GNU Make and with hbmk2.
* config/win/cygwin.mk
+ Added HB_CCPATH, HB_CCPREFIX, HB_CCPOSTFIX support.
To be in sync with rest of gcc family compiler in GNU Make
and with hbmk2.
* config/dos/djgpp.mk
+ Added HB_CCPATH, HB_CCPREFIX, HB_CCPOSTFIX support for
dos/djgpp.
! Fixed to only give error when using non-DOS build of
GNU Make on a Windows host.
; NOTE: Tried the bld process with bash.sh, and it starts
when using nativ DOS make.exe, but it fails also
with 'Bad command or file name' and 'No command name seen.'
when building hbcommon lib. Same effect when using
msys. So I conclude our djgpp make files are broken
for POSIX shells, I hope someone else can also do
tests.
* config/os2/gcc.mk
+ Added HB_CCPATH support.
To be in sync with rest of gcc family compiler in GNU Make
and with hbmk2.
* config/win/bcc.mk
! Readded dbl quotes around tlib filename. This breaks tlib
on msys shells. The real problem is caused by /P64 option.
Fix unknown, besides avoiding msys shell+bcc target
combination.
* contrib/hbfbird/Makefile
* Indented/rearranged.
* harbour/source/vm/hvm.c
* harbour/source/common/expropt2.c
* cleaned casting in date and timestamp math operations to use exactly
the same conversions in all places in compiled and HVM
* bin/postinst.bat
! Fixed to work without HB_TOP value. Regression after recent change.
* contrib/rddsql/sddmy/mysqldd.c
* contrib/rddsql/sddpg/pgsqldd.c
* contrib/hbbtree/hb_btree.h
* contrib/hbbtree/hb_btree.c
* source/vm/macro.c
* source/vm/hvm.c
* source/common/expropt2.c
* source/compiler/complex.c
! Fixed MSVC warnings.
Please review me, especially the date related ones in core.
* LONG -> long in hb_compExprReduceTimeStampPut().
; NOTE: Number of MSVC (32-bit) -W4 warnings are reduced to 280,
all of them in RDD code. (plus just few more here and there)
* utils/hbmk2/hbmk2.prg
! Fixed silly regression in 2009-08-23 15:56 which cause RTE on
startup if no comp/plat were specced.
* utils/hbmk2/hbmk2.prg
* config/wce/poccarm.mk
* config/wce/msvcarm.mk
* config/win/xcc.mk
* config/win/pocc.mk
* config/win/icc.mk
* config/win/pocc64.mk
* config/win/msvc.mk
! Changed to use '-' option prefix instead of '/' to make them
work under sh (msys) shells.
* INSTALL
+ Mentioning dosemu.
* bin/postinst.bat
+ dos shell flavour zip creation.
* config/global.mk
+ Added make platform detection.
+ Changed 'dos shell on win platform' warning to be based on make
platform.
+ Not showing 'dos make on win platform' warning when doing DJGPP
build because this is the right (and only) way in case of this
compiler.
* config/globsh.mk
+ Added command line length limit details for dos shell.
+ Added command line length limit for OS/2. Copied from other .mk file.
* config/dirsh.mk
* config/instsh.mk
- Deleted comments 'We have to use script to overcome the DOS limit of max 128 characters'
from all places.
* config/dos/djgpp.mk
+ Will now error if trying to use other than DJGPP make
to built with this compiler.
* bin/postinst.bat
+ Reworked to work better in cross-build/cross-shell scenarios.
* config/common/watcom.mk
! Use envvars to pass parameter either if shell is dos or target.
This fixes dos/watcom builds under non-dos shells.
* source/dynlib/mt/Makefile
* source/dynlib/Makefile
* config/none.mk
! Fixed the way these dynamic lib targets are skipped.
* package/mpkg_win.nsi
* Changed to also work on trees using short names.
(only 'ChangeLog' is a problem here).
Maybe it's time to rename to CHANGES. It causing too much
headache.
* bin/postinst.cmd
* Sync with .bat.
* config/global.mk
+ Exporting HB_SHELL for postinst scripts.
! Fixed HB_PKGNAME assembly for dos targets.
! Typo in comment.
* config/win/bcc.mk
! Deleted dbl quotes from tlib commands as it totally confused it
under msys shell.
* doc/bintools.txt
* config/dosecho.exe
! Restored previous version. (new one gave error on some watcom lines f.e.)
+ Added binary+source links.
* INSTALL
+ Some minor clarifications.
* 'NT system required' changed 'XP system recommended' for win hosts.
+ Added shell requirement for HB_BUILD_PKG.
% Simplified os2/watcom on linux example.
* bin/postinst.cmd
+ Some cleanup/sync with postinst.bat.
* source/Makefile
! Possible fix: Emptying DYNDIRLIST_BASE on dos targets
to save envvar space.
* source/Makefile
* config/lib.mk
+ Added gtcrs and gtsln to dynlib obj list for *nix targets.
* config/global.mk
+ Added Harbour version number to log header.
! Fixed to use short pkg name based on target platform
no host platform (=dos).
* config/bsd/gcc.mk
* config/hpux/gcc.mk
* config/darwin/icc.mk
* config/linux/icc.mk
* config/linux/sunpro.mk
* config/sunos/gcc.mk
* config/sunos/sunpro.mk
+ Added dynamic library generation to rest of targets.
; Please review, especially sunos/sunpro as it had some
.sh tricks which I couldn't replicate.
* config/bsd/global.mk
* config/hpux/global.mk
* config/darwin/global.mk
* config/linux/global.mk
* config/sunos/global.mk
! Setting DYN_PREF to 'lib'.
* config/wce/mingwarm.mk
* config/win/mingw.mk
- Disabled import libs. [They aren't necessary with mingw.]
* config/darwin/icc.mk
* Changed to use libtool instead of xiar.
(blind sync with darwin/gcc)
* config/dos/djgpp.mk
* config/common/watcom.mk
* config/win/bcc.mk
* config/os2/gcc.mk
! Fixed to use $(ECHOQUOTE) in all $(ECHO) calls.
* config/win/bcc.mk
! Fixed use fixed backslashes regardless of shell in scripts.
* config/dyn.mk
! Fixed to use $(LIB_PREF) as implib name prefix.
% Minor opt in IMP_NAME forming.
* config/rules.mk
% Readded $(HB_INC_COMPILE) to be always on cmdline, to
better balance between cmdline and envvar space. (dos)
* config/instsh.mk
% Don't display useless '! Nothing to install' message
when there was no file to install. (could happen with
mingw dynamic lib install where implib is empty because
we don't generate one).
* config/dosecho.exe
! Replaced it with a different build which is coming from
a known location, the DJGPP package, where also the other
included .exe came from. This is an inferior build, but
for our purposes it will do.
* INSTALL
* config/global.mk
* Minor.
* INSTALL
+ Improvements. (f.e. moved non-*nix sections to top of
quick instruction list as *nix users tend to know these
things already)
; NOTE: Is anyone interested in proofreading it or just plain
giving some feeback? It's great that I update this doc,
but I still had no feedback whether is it suitable for
any purpose, or useful for anyone. Anything missing,
incomprehensible, or god forbid: useful.
* INSTALL
% Further simplified quick start instructions.
- Deleted HB_COMPILER from non-*nix build cmds.
* Replaced HB_COMPILER with HB_PLATFORM on *nix platforms.
+ Replaced cross-platform scenario list with a complete
host platform/shell vs. target platform/compiler
compatibility matrix.
Feel free to correct/verify/extend it.
It's quite scary so I added it as last section.
* config/global.mk
* INSTALL
+ Added poccarm and pocc64 autodetection.
+ Added iccia64 autodetection.
+ Added watcom target platform detection on all host platforms.
+ Added djgpp autodetection under win hosts.
% Simplified INSTALL examples accordingly.
Actually it's so simple now that only 'make' should be
issue after setting up the compiler according to compiler
docs. It's very rare that HB_COMPILER or HB_PLATFORM should
be specified now on non-*nix platforms.
* Minor formatting.
- Deleted debug $(info) line.
* utils/hbmk2/hbmk2.hu_HU.po
* utils/hbmk2/hbmk2.prg
! Fixed to not initialize HB_DYN_INSTALL location to central lib
dir in case there was a plat/comp specific lib dir found.
I can't remember why this was done this way, but now it doesn't
seem to be necessary and it can just cause problems, as central
lib may contain some overlapping lib name (leftovers from other
'install' sessions) causing trouble for linkers. It also cleaner
this way.
! Added '(internal)' text to Harbour command -trace output.
! Changed harbour.exe path in internal command trace/error output
to use the same base dir as hbmk2 itself. In cross-compile
situations this is more accurate. Since harbour.exe shouldn't
necessarily exist, I've put harbour.exe path in paranthesis
to suggest it's not a real path.
; NOTE: Actually now hbmk2 autodetection code is lagging behind
GNU Make, so hbmk may need to get a bigger internal revision
sooner or later to catch up. It currently cannot switch
platforms based on detected situation, also the embedded
and normal tool installation should be merged, plus
autodetection added for *nix platforms.
* include/hbwince.h
* include/hbdef.h
! Added HB_EXPORT to some WinCE function, because they are needed
to successfully link shared final executable for non-mingw
wce compilers (poccarm and msvcarm). (in mingw we automatically
export all 'extern' functions, so this isn't a problem)
* harbour/config/global.mk
! fixed MinGW cross build autodetection
TOFIX: find_in_path_par do not work for paths separated by spaces
due to $(substpat) trick
* harbour/include/hbmacro.h
* harbour/source/macro/macro.yyc
* harbour/source/macro/macro.y
* hb_macroIdentNew() made public
* harbour/include/hbexprb.c
! use hb_macroIdentNew() instead of hb_compIdentifierNew()
in code compiled for macrocompiler
* INSTALL
* Cleaned examples.
- Arranged them by host platform
- Fixed a few examples (DJGPP on NT missed HB_PLATFORM)
- Moved 'set HB_COMPILE' to make cmdline.
- Consistency changes.
- Deleted empty 'rem' lines.
- Moved log redirection from each example to general NOTE.
- Moved %1 %2 from example to general NOTE.
- Added generic *nix section.
- Added information about possible cross-build targets.
; Hopefully they are much cleaner now. In fact most information
deal with how to setup the compiler.
* config/global.mk
+ Added link to GNU Make free book.
% path checking functions rearranged, so now it can be indented.
+ Added log header line.
! Fixed to display only the first autodetect location.
* bin/postinst.cmd
* bin/postinst.bat
* bin/hb-func.sh
* Wording: 'Generating' -> 'Making'
* 'Harbour Make (hbmk2)' -> 'hbmk2' in generated hbmk.cfg.
* utils/hbmk2/hbmk2.prg
* Wording: 'Generating' -> 'Creating'
* harbour/include/hbexprb.c
* minor modification
* harbour/source/common/expropt2.c
+ added compile time optimization for expression like ("alias")->var
and ("alias")->(<exp>)
* contrib/hbcurl/hbcurl.c
* contrib/hbcurl/hbcurls/Makefile
* contrib/hbcurl/Makefile
- contrib/hbcurl/hbcurlm.c
- Deleted two wrappers belonging to share/multi interfaces.
* Moved two legacy functions back to main source.
- Deleted hbcurlm.c which caused doubly defined symbols
when built with pocc. It was caused by some curl header
features most probably.
* config/global.mk
* Comment update in config var planning section.
* config/global.mk
+ Will now show the autodetected compiler path.
* utils/hbmk2/hbmk2.prg
! Fixed to not pass bogus resource source files to linker when
resource compiling isn't supported on given compiler (win/icc).
* config/common/watcom.mk
! Fixed double .exe extension (regression).
* contrib/hbqt/Makefile
* contrib/hbxbp/Makefile
* contrib/gtqtc/Makefile
+ Experimental changes:
Now showing reason why library was skipped.
Configuration also works a little differently:
HB_INC_QT should either be set to the header dir,
or set to 'no' which will disables any autodetection
(on *nix systems). This method replaces HB_HAVE_QT=no
setting and leaves only one setting to control QT
lib building. Experimental, probably the control
method should be further tuned.
Possible log messages:
- component not found
- component location not set
- deselected
- platform not supported
- compiler not supported
- (...) not supported
+ not necessary
NOTE: I've used 3.81 'if/else if' structure here.
% Streamlined the way darwin autodetection is solved.
(not yet tested)
* contrib/hbfbird/Makefile
* contrib/hbfimage/Makefile
% Optimized prerequisit check by using $(filter)
* config/win/icc.mk
! Changed to use xilink instead of using the compiler for
linking. This syncs it with msvc and hbmk2, and it also
fixes .dll creation.
+ Added /nologo.
* config/global.mk
* config/wce/poccarm.mk
* config/wce/msvcarm.mk
* config/win/xcc.mk
* config/win/pocc.mk
* config/win/bcc.mk
* config/win/watcom.mk
* config/win/icc.mk
* config/win/msvc.mk
* config/rules.mk
* config/os2/watcom.mk
% Optimized OBJ_DYN_POSTFIX variable to be an internal one,
it doesn't need to be set in compiler .mk file anymore.
If there is any extra C compiler flag set for dynamic mode,
it will be set automatically.
- doc/howtobsd.txt
* doc/Makefile
* INSTALL
* Essence of the content moved to INSTALL.
* config/none.mk
+ Added support to display the reason for the "skip".
* external/sqlite3/Makefile
* contrib/gtalleg/Makefile
+ Showing the reason why the libs were skipped.
Experimental. This feature needs some more design
to be consistent and helpful, it should be done while
also standardizing our config methods.
Current internal solution is also not terribly elegant,
maybe it can be converted to a function call, but there
may other solutions too.
* doc/dirstruc.txt
* Minor update.
* config/globsh.mk
! Deleted double quotes from around $(MK) command on nt
shells.
This had the side effect that GNU Make couldn't run the
executable for recursion if it wasn't in the PATH.
(so win-make.exe didn't work. Now it does.)
- doc/linux1st.txt
* doc/Makefile
* INSTALL
+ Moved linux1st.txt content to INSTALL.
* utils/hbmk2/hbmk2.prg
* config/global.mk
- Deleted some HB_ARCHITECTURE compatibility lines.
Now only legacy hbmake sources keep referencing it.
