On Thu, 5 May 2005, Richard M Kreuter wrote:
> With SBCL 0.9.0 on NetBSD/x86, there are a few failures getting
> through the test suite (all the repeatable ones in debug.impure.lisp)
> This one, from the bottom of debug.impure.lisp, I can narrow down to a
> proximate cause: non-encapsulating traces don't seem to work.
> gdb reports a SIGTRAP followed by a SIGSEGV, but otherwise isn't much
> help, as the error is occuring in Lisp-land. Any suggestions figuring
> out what's going on here or how to debug this are welcome.
The SIGTRAP is most likely from the breakpoint installed by TRACE. The
SIGSEGV is the "real one".
It's a relatively new test, so it may have been broken on NetBSD/x86
before. A good starting point would be to verify if older SBCL's exhibit
the same behaviour. If not, you can do a binary search to find the commit
that broke it, in which case fixing it will probably be a breeze.
If it was broken already in the past, then things are trickier. Juho
Snellman recently fixed non-encapsulating trace on x86-64, so he may have
some debugging hints. Juho?
-- Nikodemus Schemer: "Buddha is small, clean, and serious."
Lispnik: "Buddha is big, has hairy armpits, and laughs."