I've been entertaining myself off and on over the last few days, with
encouragement and suggestions from Christophe, trying to make his
reasonable DEFTRANSFORMs for %DATA-VECTOR-AND-INDEX work the way that
they should despite various bugs in SBCL.
There were two kinds of misbehavior involved, one of which seems to've
been solved when Christophe fixed the SB!XC version of ARRAY-HEADER-P.
The other one is still unsolved, but at least I've now boiled it down
to a reasonably tidy test case. Since I'm now unlikely to work more on
it for at least a day, I thought I'd throw it up here in case someone
else wants to look at it.
To see the bug, start a fresh SBCL, compile the file, and load it.
Christophe said that a slightly earlier version of this test case also
failed on CMU CL, so very likely the current version does too, but I
still haven't gotten CMU CL for OpenBSD so I can't conveniently test
William Harold Newman <william.newman@...>
fomenting unrest between the best and the good since 1969
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C