[armedbear-devel] Planning for abcl-1.1.0
evenson at panix.com
Sun Jul 29 08:06:32 UTC 2012
Yesterday (and today), Erik and I spent time discussing and preparing
for the engineering work necessary to get abcl-1.1.0
released. This release will obviously feature Rudi's hard work as a
reasonably stable version of AMOP that CLOSER-MOP can be patched to
reference, but we have other goals as well which I want to capture so I
now transcribe a bit about our thinking, decisions, and plans going
forward. Comments, objections, innovations, etc. solicited.
We now have [some 34 active bugs][abcl-1.1.0] in the ABCL 1.1.0
milestone that we intend to address.
We have assigned an owner to most of the bugs of either Erik or myself.
This does not mean that others shouldn't be working on the bug as well
if they have time to contribute to it, just that they should co-ordinate
to the "owner" (a quick #abcl message like "ehu: I have stuff for #19?,
do you have any patches?" would be fine to avoid duplication). Roughly,
Mark got everything dealing with abcl-contrib and Pathnames, and Erik
got everything dealing with the compiler.
We intend to have a weekly meeting on #abcl, tentatively for set for
Mondays around 2100 CET which we will review what we have gotten done,
ask questions, and generally try to move the release forward. Everyone
is welcome to contribute feedback, volunteer help, etc. as these will
be open meetings.
We plan on an abcl-1.1.1 maintenance release 30 days after abcl-1.1.0
that will fix everything that we can.
An explanation of the criteria for bugs we selected:
# Criteria for issues
In order of importance:
0. Enable CLOSER-MOP.
1. No regressions with respect to the behavior of abcl-1.0.1.
Fortunately, there only seems to be one categorized with "regression"
tag (is this really the correct situation?)
2. Load a wider range of Quicklisp systems. Fix the ANSI test-grid
problem if it is still active, using the source code distributed by
Quicklisp as the basis for a more robust system. We might not
necessarily fix the compiler in all cases, but instead implement a
fallback strategy to include source forms in our FASLs for which to pass
to EVAL for forms that we can't compile. This is under active
consideration, as it presents the single largest unknown in terms of
necessary resources to address.
3. Fix all known "ansi-conformance" problems.
4. Fix everything reported as broken by users.
In the weekly meetings, we may start "jettisoning" some issues for later
releases as time goes on.
We estimate that given just Erik's and my time, it will take roughly
three months from now to release abcl-1.1.0. If you want to make this
happen faster, we welcome all helping hands.
"A screaming comes across the sky. It has happened before, but there is
nothing to compare it to now."
More information about the armedbear-devel