R e l e a s e N o t e s +------------------------+ v3.2.2 Beta 1: -------------- A few bugfixes and a some new features. o Locale support in libc. o Implmented a different approach to OMF and weak symbols. This new hack does a pre-linking before invoking the OMF linker generating new .def if needed and an extra object file with aliases to resolv the weak symbols correctly. The old approach used when GCC_WEAKSYMS is present in the environment. Please note that due to a bug in BFD weak symbols doesn't work 100% correctly. This will be addressed when we now move on to a later version of binutils. o A couple of new functions in libc, and a few correction of existing in respect to namespace. o Addressed issues with _System, _Optlink and __stdcall. They should all be working now. o Changed GCC to link with shared libc by default. This means that in order to link with static libc one have to specify -static. It also means that -Zmtd or -Zcrtdll is no longer required for linking with shared libc and can be removed from all makescripts. o Changed EMXOMFLD defaults to ilink from Visual Age for C++ v3.6.5 or later (preferably later). The emxomfld usage output will instruct you how how to use older linkers. For VAC308 and VAC365 linkers /dbgpack is specified if debuginfo is to be generated. Emxomf checks for EMXOMFLD_TYPE=link386 when choosing debug info level. o Shipping three extra libc dlls: libc02.dbg: Unstripped version of libc02.dll. This is identical except from the presense of debug info. libc02.prf: Linked together from profiled enabled libraries using kProfiler. libc02.elh: Standard libc but linked with an external eletrical fence heap from kLibDbg. These libc02.* dlls can be used instead of libc02.dll. kLibDbg and kProfiler is available in separate zip files at same location as the rest of the Innotek GCC suite. However these external tools are provided as is. o Distribution is splitted into several zips: -core.zip: All you really need. -dbg.zip: Map files, unstripped tools and libs, and libc02.elh. -prf.zip: Profiling libraries and libc02.prf. libc.zip: The runtime libraries. (included in core.) The big zips is everything. o I'm sure I've foregotten something this time.... v3.2.2 Alpha 3: --------------- Primarily bugfixes. o Changed code generator for floating point truncations (double to int casts and such) so PM can freely mess up the fpu status. o We were using the wrong float.h in two sense. We're now using the correct one which includes all the PC defines and prototypes. o Fixed a few issues with error defines and tcpip. Note that some of the error defines have changed value in the process so we could become compatible with the tcpip stack and FreeBSD. o Corrected bugs in a number of headers. o Added more OS/2 prototypes and imports. There is still work left on the header part, o Optimized emxomf debug info conversion. o Wrapped bsdselect for BSD 4.4 header mode to implement bsdselect(). o .... what did we forget... v3.2.2 Alpha 2: --------------- A few bugfixes and a couple of new features. o The TCP/IP headers and libraries scheme have been changed. The default is the current OS/2 tcpip toolkit (BSD 4.4 based). To target the older OS/2 tcpip stack (BSD 4.3 based) you must define TCPV40HDRS before including any TCP/IP headers and make sure usr/lib/tcpipv4 is searched before usr/lib (this is to get the right libsocket). It is recommended to use the -D compiler option for the define and either the LIBRARY_PATH or the -L compiler/linker option for the library. o There are a couple of new and updated based on FreeBSD 5.1. sys\stdint.h and inttypes.h (and a few internal once) are new. sys\types.h and unistd.h are the ones most heavily updated. We're working on getting the headers and libc in a better shape standardwise to easy compatability and portability. o A infinite loop in emxomf has been fixed. o Linker error on sopen() has been corrected. o .... what did we forget... v3.2.2 Alpha 1: --------------- This is the first release from Innotek, and it contains the Innotek LIBC. In addition to the new LIBC there is a number of other major changes: o Name mangling in OMF mode have changed. Until this release emxomf used to pretty much strip one leading '_' from symbols when converting them to omf. This is now longer done chiefly in order to correctly support mangling calling conventions like __stdcall and __cdecl. This change have the following side effects: - _System and __cdecl no longer produces the same names and hence all OS/2 API which are _System must be declared as _System in order to link. - Assembly code might need an extra underscore to work, or perhaps change the prototypes to _System. - Global variables are prepended and underscore, so mixing VAC and GCC, or accessing globals from asm, take care. (MSC v6.0 does this too, so it's not any uncommon practise.) The OS/2 headers which are included are all updated to use _System. o TCPIP headers are updated a quite a bit to match the toolkit ones and to reflect the fact that the new libc does not a having common handle space for files and sockets. These changes aren't 100% completed yet, and usage probably will uncover some issues left in there. Also for the programmer there is the following changes in the apis provided: - TCPV40HDRS are supported and should work like with the toolkit headers. It will alias all functions to tcpip40_ editions (using #defines) so we get the right imports thru -llibsocket. - There is no support for sockets in select(). To do select on sockets you will have to use bsdselect() or os2_select(). bsdselect() is only available for TCPV40HDRS. For !defined(TCPV40HDRS) fd_set is in the emx/bsd format and not the OS/2 tcpip v5.0 format. - Socket IOCTLs are not supported by the ioctl() libc function because of the separate handle space for sockets. For sockets so_ioctl() and os2_ioctl() are used. os2_ioctl() is not available when TCPV40HDRS is defined. According to the tcpref in v5.0 so_ioctl() only supports BSD ioctls. - writev() and readv() also clashes with libc. For socket handles use so_writev() and so_readv(). o The prefered OMF linker is now ilink. link386 will fail with an odd error message if you tell gcc/g++ to make debug info. Set the EMXOMFLD_LINKER environment variable to value ilink in order to use ilink as OMF linker. It likely that we will change the default OMF linker to ilink in an later version of the GCC suite. o .... what did we forget...