On Thu, 2005-10-27 at 00:00 -0400, John E. Malmberg wrote: > Everything in Perl is covered under the Perl Artistic License which is > not quite the same as GPL, so mixing some code may take getting > permissions from the various authors.
Ah. I hadn't considered that. Thanks for the reminder. > GNV ships with GPL source so if Ruby is GPL, then there is no problem > with the interchange. Ruby is dual-licensed GPL + Ruby's own license. > Usually I am finding that it is more useful to learn from the example > than to snip the code. My pipe change to bash looks nothing like the > code in Perl, even though it was what gave me the idea. OK. I'll continue along the same vein, then. > The config.guess file from the current coreutils (5.2.1) now knows about > current OpenVMS. And that is one of the key things to for getting > Configure. scripts to run on bash. I'm working now with someone else outside of our company who recently expressed an interest in the VMS Ruby port. He evaluated the autoconf for VMS package and decided for now it is just simpler to write a mkconfig.com that generates our rbconfig.rb tailored for our environment, and which all other ruby components expect to exist. One of our objections to depending on GNV + bash at this point is that both of us gave it (admittedly limited) trials and at the time we looked, it was not yet complete or stable enough to be worth further investigation. So unless things have changed quite a bit since we each last looked, in the interest of minimizing the Ruby build dependencies and also of not depending on something that isn't yet ready for prime time, we'll continue to use our hand-rolled configuration approach. So long as the VMS Ruby port is not in widespread use, and is still rapidly changing, we expect this will work out. We haven't ruled out switching build systems later, though. > For some packages, just using that is enough to get Configure. to run > the first time. The RWMBX hang comes from the second Configure. run > because it passes the results of the first run through "sed", and it is > too big to fit in a mailbox. Delete the stuff left over from a previous > configure, and then it works again. Sounds a bit dodgy. > I have not had the time to see what the Perl Configure. script would do > under my modified bash. If it works, I'd be impressed. It might influence our build system decision. > I am also trying to get more of the routines into the OpenVMS CRTL, but > there is usually a long lead time so that all the testing can be > complete when the Operating System version ships. Yes. That was my impression of the process. I was quite discouraged when I ran up against a wall with differences in Ruby's assumptions about the equivalence of sockets and other fds vs. what the VMS CRTL actually implements. I ended up with the impression that OS support for a more "Unixy" handling of fds was on the never-never plan. Ben
