That all sounds reasonable. How about this. Mail me directly with a list of
the exact errors you got and where. I'll look at those and we'll see what
we can do. We definitely don't want to start putting in conditional code
per environment. But, the other stuff I'll see what I can do to clean those
up in a way that would make everyone happy. When I get that taken care of,
you can check out the latest code and make sure that your stuff is still
happy and functional. If it is, you can mail us the stuff and we'll
integrate it into the code base.
One question... What is the size of wchar_t on your 64 bit system? The
parser code should work well with it being set to either 16 or 32 bits, but
might get freaky if a character is 64 bits.
----------------------------------------
Dean Roddey
Software Weenie
IBM Center for Java Technology - Silicon Valley
[EMAIL PROTECTED]
Gary Gale x58814 GB- COM/Q/6th <[EMAIL PROTECTED]> on 03/15/2000
01:26:02 AM
Please respond to [EMAIL PROTECTED]
To: [EMAIL PROTECTED]
cc:
Subject: Re: Xerces-C on Tru64 - An Update and Some Questions
>>Before submitting the changes, how about just giving everyone a
discussion
>>of what was required to make it work. This would give us a chance to
>>comment on what you've done and, if we feel something needs to be changed
>>before accepting it, that you can make those changes and test them out
>>before submitting.
Sorry for the lack of info in the original post. I was none too sure what
the
protocol was and didn't want to bombard the list with a very lengthy (and
probably unwanted) initial posting. (Takes a deep breath, here goes ...)
The Tru64 port is currently being hosted on an AlphaServer 4100 5/400
running
Digital Unix/Compaq Tru64 V4.0f (Rev 1229) with 2048Mb of physical memory.
The latest version of Xerces-C was used (1.1.0).
The build environment consists of :-
Compaq C++ V6.2-024 for Digital Unix V40.f
GNU Make V3.77
GNU autoconf V2.13
GNU m4 V1.4
1) Compiler Support
Created src/util/Compilers/DECCXXDefs.hpp and DECCXXDefs.cpp
Standard typedefs for XMLCh et al, plus implementation of stricmp &
strnicmp.
2) Platform Support
Created src/util/Platform/Tru64/Tru64Defs.hpp and Tru64PlatformUtils.cpp
Defines ENDIANMODE_LITTLE (Tru64Defs.hpp) and the required Platform
utilities. I based these on the Solaris version; the only major change
required was to base XMLPlatformUtils::getCurrentMillis () on the use
of clock_gettime (CLOCK_REALTIME, &) rather than ftime which requires
_XOPEN_SOURCE_EXTENDED to be defined and thus messes up thread
support.
3) Overall XML and Autosense Support
Modified src/util/XML4CDefs.hpp
Includes Tru64Defs.hpp and DECCXXDefs.hpp if XML_TRU64 is defined.
Modified src/util/AutoSense.hpp
Defines XML_TRU64 if __alpha and __osf__ are defined.
Defines XML_DECCXX is XML_TRU64 and __DECCXX are defined.
4) Makefile Changes
Modified src/util/Platforms/Makefile.in
Added macro definitions for SUBMODULE, CPP_PUBHEADERS and CPP_OBJECTS
if the PLATFORM macro is set to TRU64
Modified src/util/Compilers/Makefile.in
Added DECCXXDefs.hpp to CPP_PUBHEADERS macro
Added macro definition for CPP_OBJECTS if CXX macro is set to cxx
and the PLATFORM macro is set to TRU64.
Modified src/Makefile.incl
Added compiler options if the PLATFORM macro is set to TRU64. The main
additional compiler switch required was to create a (hidden) template
instantiation directory ($XERCESCROOT/.cxx_repository using the -ptr
switch. Other macro definition are as for Solaris.
Modified src/configure.in
Added trapping of *-*-osf* and setting the PLATFORM macro to TRU64
when checking for the host type.
Modified src/runConfigure
Added support for -p tru64 and -x cxx.
5) Building
With the above changes in place I built the system as follows
$ make distclean
$ autoconf
$ ./runConfigure -ptru64 -ccc -xcxx -minmem -nfileonly -tnative
$ make
$ make install
6) Samples
Modified samples/runConfigure, configure.in and Makefile.incl.
As discussed above.
>>On the subject of the mixed signed/unsigned stuff, yes those are being
>>removed as time goes by. I think that most of them are gone in the
current
>>code base. And, BTW, what code base have you been working against? We'd
>>definitely want any submitted changes to be against the most recent
code
>>base.
For the code base stuff see right at the top of this post.
These are the source/header files that DEC CXX got stroppy about.
Signed/unsigned integer conversions and pointless comparisons
IConvTransService.cpp
XMLScanner2.cpp
XMLValidator.cpp
DOMString.cpp
NodeVector.cpp
TextImpl.cpp
CMNode.hpp
CMLeaf.hpp
DFAContentModel.cpp
Unreachable statements:
XMLReader.cpp
XMLScanner2.cpp
NamedNodeMapImpl.cpp
DTDValidator2.cpp
Let me know what you think.
All the best
Gary
--
Gary Gale Mail: [EMAIL PROTECTED]
UK Server Group Phone: +44 (0) 171 542 8814
Factiva, A Dow Jones & Reuters Company Web: www.factiva.com
