| Re: [stellation-res] Scripts in Stellation |
On Sun, 2002-09-15 at 22:42, Jonathan Gossage wrote: > Scripts are used in a number of places in Stellation, both as application > launchers and as test harnesses. Currently these scripts are implemented as > shell scripts, either for bash or sh. This presents a problem on Windows > platforms even those where Cygwin is installed. Basically, any *nix script > that deals with file manipulation needs modification in Cygwin to handle > translation between DOS and *nix paths using Cygwin specific extensions. > While it would be possible to modify these scripts so that they could run > native in a *nix environment or in Cygwin, I do not believe that we should > be relying on Cygwin as a Windows environent over the long term. I'd really like to replace the test harnesses with unit tests, which would take care of that problem. I know we've been saying that for a while; there just isn't enough time for all the things that need doing. But the script tests are inaccurate, hard to control, are are non-portable. They should be replaced with a rather larger set of unit tests that test specific functions more carefully. For the other scripts... They're mostly pretty trivial. Providing overlapping shell scripts and .bat files is easier, I think, than writing fully portable perl or python scripts. Besides which, not all windows users have perl or python, and I'm already concerned that Stellation's dependency set is overcomplicated. Finally, for launchers, I think that in the long term, a full plugin architecture is the right way to go. Instead of the shell script launcher that we currently use, I'd like to wind up using the Eclipse launcher. Unfortunately, it's another time issue: no one has the time right now to make that work. > I would like to propose that the current scripts be rewritten in a platform > independent scripting language such as Perl or Python. If this is done right > we will have one set of scripts that can be used on all platforms, perhaps > with some conditional platform dependent code in them. Writing that kind of fully portable code is tricky - given the triviality of most of the scripts, probably more tricky than maintaining parallel scripts. -Mark -- Mark Craig Chu-Carroll, IBM T.J. Watson Research Center *** The Stellation project: Advanced SCM for Collaboration *** http://www.eclipse.org/stellation *** Work Email: mcc@xxxxxxxxxxxxxx ------- Personal Email: markcc@xxxxxxxxxxx
Attachment:
signature.asc
Description: This is a digitally signed message part