Platform notes


Solaris - gcc 2.8 has a bug in the optimizer, causing it to fail with Internal compiler error (usually while compiling custom.cpp, but there have been other cases). Removing "-O3" from the compile flags works around this.


MacOS X (rhapsody) - May 1999 - Requires autoconf 1.3 to recognize the new system. Configure does not find a working "mmap(2)" - (why?) Libtool cannot build shared libs (perhaps because gcc is still at version 2.7.2.1?) - static passes all tests.

MacOS X 10.0.3 - May 2001 - Fails to recognize this platform (powerpc-apple-darwin1.3.3), fixed by doing:

    cd ../unix
    glibtoolize -f -c
    cd ../builds

The "./configure" works. This version builds static and shared libs (with the proper ".dylib" suffix), but "regress" is confused by the mac vs. unix context and produces "tests\*.txt" files (i.e. not in the "tests/" directory). Also, "myio" fails to build, probably due to a difference in shared libs.

MacOS X 10.1.1 - Nov 2001 - now uses MK, as part of "Address Services". --jcw

MacOS X 10.1.1 - MK 2.4.2 builds cleanly and passes all tests. --jcw

Whoops: no it don't. Tests s13 and s21 hang. This appears to be caused by some unfortunate interaction between mmap and I/O. I do have a workaround for now (which does not disable mmap). --jcw


Symantec C++ - A note to say that I am building Metakit for DOS and DOS32 using Symantec C++. See also STL and Metakit. Please contact me if you are interested in this. -- John Fletcher


Salford C++ I am building Metakit 2.33 and E4Graph using this compiler, in order to link them to FORTRAN and be able to call storage from FORTRAN. The problems due to the fussiness of the compiler have now been resolved and I am now working with the latest versions of Metakit and E4Graph. -- John Fletcher


GCC on mingw32 I am also building Metakit 2.4.1 and E4Graph 1.0a4 using GCC 2.95.2 under mingw32. The motivation for this is to be able to integrate the Lambda Library. I have also been able to build DLL's for this system. --John Fletcher


BeOS - John Bushakra reports having succesfully built MK 2.01 on BeOS (4.52, I think). It took some tweaks to #define's. His archive has been uploaded to https://www.equi4.com/pub/mk/mk-2.01/ --jcw


IBM s390 - MK 2.01 reportedly compiles fine under Linux 2.2 on IBM s390, and passes all tests (Nov 2001) --jcw


FreeBSD - R. Nelson reports that "the latest tclkit now builds from the metakit port on FreeBSD 4.4 from the ports tree as an option from /etc/make.conf", in other words: MK is now in the FreeBSD 4.4 distribution --jcw

FreeBSD 4.4 - MK 2.4.2 builds cleanly and passes all tests. --jcw


Linux/PPC 2000 Q4 - Dec 2001 - MK 2.4.2 builds cleanly (gcc 2.95.3), tests m01 and m03 hang. --jcw