# # MACHINE -- operating system specific information # nmh is known to compile on the following platforms (save the exceptions noted below), using an ANSI C compiler, such as gcc. AIX 4.1.5.0.01 FreeBSD IRIX 6.5 Linux 2.2, 2.3, 2.4 (glibc 2.1, glibc 2.2) Mac OS X Public Beta NetBSD 1.4.2 OpenBSD Solaris 7 and 8 (sparc,x86) SunOS 4.1 Known Compilation problems: -------------------------------------- FreeBSD: OpenBSD: NetBSD: Some BSD4.4 machines have problems when running nmh's configure script. They will be unable to find the location of vi and sendmail. This is due to POSIX features (breakage?) in the shell sh. The solution is to run the configure script under the shell `bash': % bash configure -------------------------------------- Mac OS X/Rhapsody 5: Version 5.3 at least has the same sh/bash bug as the *BSD systems above. This appears to be fixed in 5.5. Will not compile correctly unless you configure with the --enable-debug option. It appears to find conflicts in the headers only when debugging is disabled. With debugging enabled, it compiles and runs happily. -------------------------------------- HPUX: Lots of problems have been reported with using HPUX `cc'. In particular, problems with `scan' giving incorrect dates (everything is 01/00). It is highly recommended that you use `gcc' instead. Also, new versions of HPUX (10.20?) will core dump in `scan' because of some workaround code in zotnet/tws/lexstring.c. This workaround is needed for older versions of HPUX, but causes problems on newer versions. The solution is the added line (minus our indentation): #undef hpux after line 15 of the file zotnet/tws/lexstring.c. -------------------------------------- Irix (SGI): Irix make is notoriously buggy. If you're using it, you should "touch config.h.in" before configuring to prevent a problem where it tries to rebuild targets that shouldn't be rebuilt. (Alternately, you can just use GNU make instead of Irix make.) -------------------------------------- Linux: The configuration script does a test to discover if your vi is broken (if it reports non-zero exit codes on certain pseudo-errors). This test will hang if the program `ex' on your system is a link to the vi clone `vile'. The workaround is to replace the command ex as a link to another vi clone such as nvi or elvis. -------------------------------------- Solaris: With --enable-debug you'll see a lot of warnings. This is even worse when compiling using the Sun Workshop compiler since it issues a warning for every instance of a problem instead of summarizing them. The main one concerns arrays with an index of type char. This is ok. The array itself is a hash of chars, so the array size and the type match. There isn't another safe and portable way to do this at the moment. An explicit cast would get rid of the warnings, but I think it's better to leave it complaining for now until we come up with a better solution. The whole thing is probablly going to be chucked with UTC-8 support anyway. Other than the warnings, it builds ok. -------------------------------------- SunOS 4.1.1/4.1.3/4.1.4: You can't use the C compiler that comes with SunOS 4 since it isn't ANSI C. But nmh builds just fine with gcc. With --enable-debug you will see a lot of warnings. --------------------------------------