[Ecls-list] 10.3.1 slime issues

[Tobias C. Rittweiler]

Raffael Cavallaro <raffaelcavallaro at mac.com>
writes:

> On Mar 7, 2010, at 6:35 PM, Tobias C. Rittweiler wrote:
>
>> Hmm. Instead of what I said first, try `q' to quit the Emacs backtrace,
>> then enter (quit) on the *inferior-lisp*, and try M-x slime again. It's
>> a race condition at place here.
>
> infinite loop  (i.e, just get the same debug backtrace over and over)
>
>
> > 
> > I do not know why ECL segfaults (which is what you by crash, right?)
>
> emacs crashes (and I've tried about 6 different builds of emacs, both carbon and cocoa, and built myself form source).

Ok, maybe it's something else.

Could you try M-x trace-function-background RET slime-timer-call RET RET
M-x slime?

Then when you're in the Emacs backtrace, switch to the buffer
*trace-output* and paste it along the content of the backtrace buffer.


> > Alternatively, try to setup a swank-server manually, and slime-connect
> > from Emacs. If you get a SEGV by that, too, try to attach gdb before
> > connecting from Emacs so you can get at a backtrace.
>
> How does one set up a swank-server manually?

cd ~/src/slime/
ecl
(load "swank-loader")
(swank-loader:init)
(swank:create-server)

(Now optionally, attach gdb. You can get the pid by (SI:GETPID), but you
have to do it before creating the server because the server will occupy
the main thread. 

In gdb you may have to ignore some signals, on Linux it's SIGPWR and
SIGXCPU which are used by the boehm-gc; do not know about Mac, but
that's Juanjo's development platform, too, so he'll probably know.

Also build ECL with --with-debug-cflags so the C backtraces will be nice.)

In Emacs: M-x slime-connect RET RET RET


> Should I attach gdb to emacs or ecl (or both)?

ECL.

HTH,

  -T.