Menu
▾
▴
Re: [Sbcl-devel] "(quit)" in the LDB
|
From: Nathan B. <na...@ac...> - 2008-07-18 15:46:07
|
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Nikodemus Siivola wrote:
<blockquote
cite="mid:633...@ma..."
type="cite">
<pre wrap="">On Thu, Jul 17, 2008 at 8:11 PM, Nathan Bird <a class="moz-txt-link-rfc2396E" href="mailto:na...@ac..."><na...@ac...></a> wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Nathan Bird wrote:
</pre>
<blockquote type="cite">
<pre wrap="">The actual error was:
* fatal error encountered in SBCL pid 14177(tid 3072904096):
maximum interrupt nesting depth (1024) exceeded
...
fatal error encountered in SBCL pid 14177(tid 3085379264):
no size function for object at 0x09f610a0 (widetag 0xa0)
I wasn't paying attention to when it occured, just checked back and
found it was in the LDB. 1.0.18 on linux x86 built from src.
</pre>
</blockquote>
<pre wrap="">I found what triggered that, when I had rebuilt/saved the core file sbcl
was started with in another process.
</pre>
</blockquote>
<pre wrap=""><!---->
I'm sorry, but I am not sure what you mean. Can you provide the exact
sequence of
steps?
Cheers,
-- Nikodemus
</pre>
</blockquote>
1. sbcl --eval "(progn (asdf:oos 'asdf:load-op 'swank)
(save-lisp-and-die
\"/tmp/sbcl-swank.core\"))" && cp /tmp/sbcl-swank.core .<br>
2. sbcl --core sbcl-swank.core<br>
3 (separate shell). sbcl --eval "(progn (asdf:oos 'asdf:load-op 'swank)
(save-lisp-and-die \"/tmp/sbcl-swank.core\"))" && cp
/tmp/sbcl-swank.core .<br>
4. Try and evaluate something in #2's sbcl-- it hangs.<br>
5. C-c --I get the interactive-interrupt here, from which I can abort
and it works. A friend tried this(same server/setup) and it becomes
totally unresponsive (kill -9 required)<br>
<br>
If, as step 2a, I start swank and connect from emacs, step 4 gives me
the LDB with "maximum interrupt nesting depth (1024) exceeded".<br>
<br>
If I generate the cores to the current directory instead of /tmp then
copy, there aren't any problems. My best guess is the file copy is
generating a signal from the OS which sbcl isn't quite handling (the
hang in step 4).<br>
<br>
Debian etch, x86, 2 proc with hyperthreading<br>
kernel: 2.6.15-1-686-smp<br>
sbcl 1.0.18 built from src with sbcl 1.0.17 theading enabled.<br>
<br>
I do all of this to build up a standard core with ~20 libs in it;
occasionally I rebuild that core to get updates. I've been doing this
process for a while and I believe the fault is recent, perhaps 1.0.18
only.<br>
<br>
Let me know if there is something else that might help. In the meantime
I've been retraining myself to hit C-d in the LDB instead of (quit).<br>
<br>
</body>
</html>
|