Garrett D'Amore wrote: > Roland Mainz wrote: > > Erm... two items: > > 1. B72 integrated ksh93 version "s-" (which was an "alpha" version, too) > > - and for the first attempt it was AFAIK quite good. And the upcoming > > ksh93t- will be "better" since we learned from our mistakes. > > 2. Looking at the number of bugs open in bugster and other problems I > > think we can't wait much longer. ksh93t+ will likely not be available > > before mid-2009. We're simply running out of time - that's why we > > invested the whole last month with testing ksh93t- and making sure it is > > fully functional and bug-free. > > It sounds like you've done some good work mitigating the risks. > > I do have two questions though. > > 1) What criteria does the upstream source use before deciding a version > can drop the "-"?
The switch from '-' to '' (and later '+') AFAIK means we try to avoid adding new features. However for most existing scripts the new features won't be a problem - the major change was the switch from ksh88 to ksh93 where David Korn explicitly broke backwards-compatibilty to get rid of design bugs in ksh88 (AFAIK at some point he learned from the original mistakes and made a clear cut and started from scratch with ksh93). BTW: ksh93 means means "ksh based on the spec from 1993" - and since 1993 scripts based on this spec are usually guranteed to run (thinking about it... "enum" may really be the first new language-specific builtin since 1993... however I prefer to hide behind http://www.opensolaris.org/os/project/shell/shellstyle/#do_not_reserved_keywords_for_function_names). > And, perhaps more importantly, why is the testing > you've been doing not sufficient to warrant that? Erm... what do you mean with that ? > 2) What kinds of changes other than bug fixes can we expect between the > "-" release, and the version without the "-"? As long as it's '-' we may add features... when the '-' is gone we may try to avoid adding features and when the tree is marked as '+' we really try it hard to avoid adding features hard (exception may be urgend issues which cannot be solved otherwise (or in the case of ksh93-integration explicitly requested backports of newer features)) ... ... beyond that general statement then next big things are: - libshell API stabilisation and find a consumer for an ARC contract to test whether the API is complete - thread support, e.g $ ksh93 -c 'function mythr { sleep 100 ; } ; builtin pthread_create ; integer tid ; pthread_create -L -f mythr -t tid ; wait -T ${tid} ; exit 0' # ... but those two items won't create new builtins like "enum" > Put another way, I'm more concerned about incompatible changes impacting > scripts, than I am about quality (given that it seems like you've done a > lot to ensure that quality concerns are addressed.) See above... ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)