Ahoy,
I checked in the change we discussed to fetch more backtrace frames by default, and use the '--more--' token to try and avoid shoving our "SWANK"-related frames in your face. I did it with a regexp on the elisp side.
One thing to look out for is that SBCL on PPC gets an error trying to format certain backtrace frames, and since this change generally causes more frames to get formatted it can provoke the problem. The only trouble case seen so far is doing "(/ 1 0)" in the REPL.
Dan's looking into it. Meanwhile, if it causes you grief you can bung (setq swank::*sldb-initial-frames* 1) into your ~/.swank.lisp to get the old behaviour back.
Cheers, Luke
Luke Gorrie luke@bluetail.com writes:
I checked in the change we discussed to fetch more backtrace frames by default, and use the '--more--' token to try and avoid shoving our "SWANK"-related frames in your face. I did it with a regexp on the elisp side.
niiice!
one nit: the sldb-show-source command (bound to "v" in sldb) sometimes can cause errors (if it's uncommon, sorry. I'm using the latest CMU CL snapshot currently, which might have something to do with it, perhaps).
like this one: 'Bogus debug function: #<Bogus-Debug-Function "call_into_lisp+#x8C [#x8054C3C] /user/cmm/usr/linux/bin/lisp"> [Condition of type SIMPLE-ERROR]'
I wonder whether it would be a good idea to put '(ignore-errors ...)' in some appropriate place?
also, upon choosing the ABORT restart in the inner backtrace caused by the above, the frame cursor lands on the innermost frame, instead of the frame I was trying to look up source of.
but let me add, lest anybody gets the wrong impression: niiice!
Michael Livshin usenet@cmm.kakpryg.net writes:
one nit: the sldb-show-source command (bound to "v" in sldb) sometimes can cause errors (if it's uncommon, sorry. I'm using the latest CMU CL snapshot currently, which might have something to do with it, perhaps).
like this one: 'Bogus debug function: #<Bogus-Debug-Function "call_into_lisp+#x8C [#x8054C3C] /user/cmm/usr/linux/bin/lisp"> [Condition of type SIMPLE-ERROR]'
I wonder whether it would be a good idea to put '(ignore-errors ...)' in some appropriate place?
Fixed. We already had a flag *debug-definition-finding* in the CMU backend saying whether to debug these errors, but it was (until now) enabled by default.
-
Luke Gorrie -
Michael Livshin