rkrebs@archie.dsd.es.com (Randall Krebs) (01/13/91)
I just compiled perl patchlevel 44 on a sun3, a sun4, and a mips computer. Here are the problems I ran into. When applying patch 42, there was a minor problem with the file lib/ctime.pl. The patch file wanted to see the following on line #13: ;# $Date = do ctime(time); But patchlevel 41 (or before) had already changed this line to: ;# $Date = &ctime(time); which is what patch 42 wanted to change it to. Did that make sense? I fixed it by changing "ctime.pl" to the old line and patching the file with: cd lib; patch ctime.pl < ctime.pl.rej The sun3 and sun4 went pretty smoothly. We're still living in the past with SunOS 4.0.3. Configure made all the right guesses and the test ran without error on the first try. The mips wasn't so easy. I tend to prefer the BSD compilation environment on the mips. (For those of you that don't have your own mips box to kick around, there are native compilation environments and foreign compilation environments. Native is SYSV and foreign is BSD or POSIX.) Configure likes to browse through /usr/lib and /usr/include, but the real information is in /bsd43/usr/lib and /bsd43/usr/include. This makes Configure guess wrong about "i_dirent" and "d_pwage" in config.sh. I simply changed them from 'define' to 'undef'. Also, the mips compiler hates the way that perl uses volatile. I'm not sure why, but I changed "d_volatile" to 'undef', too. We're running RISC/os 4.50. If it's a compiler problem, maybe it will go away someday. I'd like to hear from anybody compiling perl with volatile defined. The test "op.stat" fails on subtests 36 and 37. I'm pretty sure that this is due to a RISC/os bug when accessing /dev/tty. Other than that, the other tests run successfully on the mips. Here are a couple of enhancement requests for perl's configuration and installation procedures: 1) Make the installation dependent on a make variable called DESTDIR. This is how X is distributed and it makes some things very convenient. Everywhere that a destination directory is referenced, it is preceded by the symbol DESTDIR. For instance, if I need to test the installation of perl without ruining the previously installed version, I could run: make DESTDIR=/tmp/test install Also, it would be nice if perl would create all the directories needed in the installation path (e.g. /tmp/test/usr/local/...) 2) Allow some options to have Configure look in non-standard places (e.g. /bsd43/usr/lib/libc.a) for libraries and header files. Mips' schizophrenic development environment is kinda twisted, but anybody that can make Configure work for Eunice could probably handle mips, too. And lastly, 3) "h2ph" ought to take command line arguments to override the source and destination directories. I want to be able to do something like: h2ph -d /tmp/test/usr/local/lib/perl /bsd43/usr/include/*.h And by the way. Thanks for the nice perl. I'm definitely a perl-junkie and you're definitely one of my net.heroes, Larry. randall. -- Randall S. Krebs | "Friends don't let friends drive BMW's" (rkrebs@dsd.es.com) | Evans & Sutherland Computer Corp. | - Harvey Shine Salt Lake City, Utah (Where?) |
meissner@osf.org (Michael Meissner) (01/13/91)
In article <1991Jan13.014908.3144@dsd.es.com> rkrebs@archie.dsd.es.com (Randall Krebs) writes: | Also, the mips compiler hates the way that perl uses volatile. I'm | not sure why, but I changed "d_volatile" to 'undef', too. We're | running RISC/os 4.50. If it's a compiler problem, maybe it will go | away someday. I'd like to hear from anybody compiling perl with | volatile defined. I do (on a DECstation), but then I use GCC (with my patches :-) to compile it. I seem to remember that the MIPS people saying that revision 2.20 of the compiler suite would finally be ANSI complaint.... -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Considering the flames and intolerance, shouldn't USENET be spelled ABUSENET?