[slime-devel] Re: Allegro 6.2 problem

[Luke Gorrie]

Peter Seibel <peter at javamonkey.com> writes:

> Luke Gorrie <luke at bluetail.com> writes:
>
>> Peter Seibel <peter at javamonkey.com> writes:
>>
>>> CL-USER> (compile-file "test-debug.lisp")
>>> Error: Received signal number 11 (Segmentation violation)
>>>   [condition type: SYNCHRONOUS-OPERATING-SYSTEM-SIGNAL]
>>>
>>> Restart actions (select using :continue):
>>>  0: Prompt for a different filename to compile
>>>  1: Abort handling SLIME request.
>>>  2: Abort entirely from this process.
>>> [Current process: repl-thread]
>>> [1] CL-USER(1): 
>>>
>>> Typing Allegro commands (e.g. :cont 1) doesn't seem to do anything.
>>> C-c C-c gets me into sldb where I can use one of the ABORT restarts to
>>> get back to the main REPL. But this seems wrong.
>>
>> I find it troubling that so many simple errors like this cause SIGSEGV
>> in ACL, but that's another story.
>>
>> Can you try again from CVS? From the ChangeLog I think that Brian
>> Downing has probably fixed this.
>
> Nope. His fix fixed my other problem (with the interactive restart)
> but not this one.

Er, I hadn't realised that the segfaults were our fault :-)

The segfault is triggered in SLDB when we call `debug:output-frame' to
format a particular stack frame for the backtrace. I don't know what's
special about this frame, but poking around in the TTY debugger I
noticed that `debug:frame-visible-p' returns NIL for it. So, as a
pretty lame workaround, I changed the backtrace to filter out all
frames that aren't visible-p.

Don't know if this really solved the problem, so please give a shout
if there are any more of these segfaults.

-Luke