[slime-devel] Debugging "Lisp connection closed unexpectedly" events?
I'm connecting with latest SLIME CVS (as of tonight) to SBCL 1.0.17 (Fedora 8). I'm trying to use Drakma to snarf the home page of LispForum with (drakma:http-request "http://www.lispforum.com/") Oddly, when I do so, I get: "Lisp connection closed unexpectedly: connection broken by remote peer" This doesn't happen when I try other web pages, however. With others, I get the text of the web page back as a string. Oddly, it does *not* happen when connecting to the same page when running SBCL in a term window as opposed to SLIME. Any suggestions for where to start looking? My gut says that there is a particular character or something like that in the page that SWANK can't return over the channel to SLIME/Emacs. I'm not sure how to debug it, though, since SWANK kills the connection with extreme prejudice. Does SWANK create any sort of log file anywhere, or can I enable it to do so? Thanks, -- Dave
Dave Roberts <ldave@droberts.com> writes:
Any suggestions for where to start looking? My gut says that there is a particular character or something like that in the page that SWANK can't return over the channel to SLIME/Emacs. I'm not sure how to debug it, though, since SWANK kills the connection with extreme prejudice. Does SWANK create any sort of log file anywhere, or can I enable it to do so?
Thanks,
Set SWANK:*LOG-EVENTS* to T for information from the swank side, and look into the buffer *slime-events* for information from the slime side. -T.
On Tue, 2008-08-05 at 08:57 +0200, Tobias C. Rittweiler wrote:
Dave Roberts <ldave@droberts.com> writes:
Any suggestions for where to start looking? My gut says that there is a particular character or something like that in the page that SWANK can't return over the channel to SLIME/Emacs. I'm not sure how to debug it, though, since SWANK kills the connection with extreme prejudice. Does SWANK create any sort of log file anywhere, or can I enable it to do so?
Thanks,
Set SWANK:*LOG-EVENTS* to T for information from the swank side, and look into the buffer *slime-events* for information from the slime side.
-T.
Thanks. There was nothing in *slime-events* at all. The command is there, but no output or diagnostics. Where does SWANK log things when *log-events* is set to T? It looks like swank:*log-output* is set to standard error, but I'm not sure what that points to when the SBCL instance is started using M-x slime in Emacs. -- Dave
Dave Roberts <ldave@droberts.com> writes:
Thanks. There was nothing in *slime-events* at all. The command is there, but no output or diagnostics. Where does SWANK log things when *log-events* is set to T? It looks like swank:*log-output* is set to standard error, but I'm not sure what that points to when the SBCL instance is started using M-x slime in Emacs.
It's in *inferior-lisp*. -T.
On Tue, 2008-08-05 at 09:19 +0200, Tobias C. Rittweiler wrote:
Dave Roberts <ldave@droberts.com> writes:
Thanks. There was nothing in *slime-events* at all. The command is there, but no output or diagnostics. Where does SWANK log things when *log-events* is set to T? It looks like swank:*log-output* is set to standard error, but I'm not sure what that points to when the SBCL instance is started using M-x slime in Emacs.
It's in *inferior-lisp*.
Ah, gotcha. Thanks. It looks like SBCL threw a SB-INT:STREAM-ENCODING-ERROR. The last bit of the output in *inferior-lisp* is: ...[lots of HTML output snipped]... </body> </html>\" " :repl-result) ;; Event history end. ;; Backtrace: ;; Connection to Emacs lost. [ ;; condition: Unable to display error condition: #<SB-SYS:FD-STREAM for "a socket" {B28D9A9}> is closed. ;; type: SB-INT:STREAM-ENCODING-ERROR ;; encoding: NIL style: :SPAWN dedicated: NIL] Any idea how I should interpret that? -- Dave
Dave Roberts <ldave@droberts.com> writes:
On Tue, 2008-08-05 at 09:19 +0200, Tobias C. Rittweiler wrote:
Dave Roberts <ldave@droberts.com> writes:
Thanks. There was nothing in *slime-events* at all. The command is there, but no output or diagnostics. Where does SWANK log things when *log-events* is set to T? It looks like swank:*log-output* is set to standard error, but I'm not sure what that points to when the SBCL instance is started using M-x slime in Emacs.
It's in *inferior-lisp*.
Ah, gotcha. Thanks. It looks like SBCL threw a SB-INT:STREAM-ENCODING-ERROR. The last bit of the output in *inferior-lisp* is:
...[lots of HTML output snipped]... </body> </html>\" " :repl-result) ;; Event history end. ;; Backtrace: ;; Connection to Emacs lost. [ ;; condition: Unable to display error condition: #<SB-SYS:FD-STREAM for "a socket" {B28D9A9}> is closed. ;; type: SB-INT:STREAM-ENCODING-ERROR ;; encoding: NIL style: :SPAWN dedicated: NIL]
Any idea how I should interpret that?
Perhaps you have to enable unicode: Add ':coding-system utf-8-unix' to the sbcl entry in your `slime-lisp-implementations', and set `slime-net-coding-system' to `utf-8-unix'. -T.
On Tue, 2008-08-05 at 09:48 +0200, Tobias C. Rittweiler wrote:
Perhaps you have to enable unicode:
Add ':coding-system utf-8-unix' to the sbcl entry in your `slime-lisp-implementations', and set `slime-net-coding-system' to `utf-8-unix'.
Okay, I'll try that. Thanks! -- Dave
* Dave Roberts [2008-08-05 08:51+0200] writes:
Any suggestions for where to start looking? My gut says that there is a particular character or something like that in the page that SWANK can't return over the channel to SLIME/Emacs. I'm not sure how to debug it, though, since SWANK kills the connection with extreme prejudice. Does SWANK create any sort of log file anywhere, or can I enable it to do so?
Yes, sounds like a character encoding problem. Maybe using utf-8 would solve it. There should be some error messages in the *inferior-lisp* buffer. This includes the condition, (sometimes) the backtrace, and the last few exchanged messages. Non-ascii chars are escaped with \xCODE. If there's nothing in the *inferior-lisp* buffer, try to look at the *slime-events* buffer or set swank:*log-events* to t. Helmut.
On Tue, 2008-08-05 at 09:12 +0200, Helmut Eller wrote:
* Dave Roberts [2008-08-05 08:51+0200] writes:
Any suggestions for where to start looking? My gut says that there is a particular character or something like that in the page that SWANK can't return over the channel to SLIME/Emacs. I'm not sure how to debug it, though, since SWANK kills the connection with extreme prejudice. Does SWANK create any sort of log file anywhere, or can I enable it to do so?
Yes, sounds like a character encoding problem. Maybe using utf-8 would solve it.
There should be some error messages in the *inferior-lisp* buffer. This includes the condition, (sometimes) the backtrace, and the last few exchanged messages. Non-ascii chars are escaped with \xCODE.
If there's nothing in the *inferior-lisp* buffer, try to look at the *slime-events* buffer or set swank:*log-events* to t.
Yup, in *inferior-lisp*, I see \x2026 characters in the string it was trying to return. How would I use UTF-8? Is the problem on the Emacs side or the SWANK side? (and as an aside, if SWANK can return all the logging info which ends up in *inferior-lisp*, how come it chokes when trying to return the same data as part of the standard REPL protocol?) -- Dave
participants (3)
-
Dave Roberts -
Helmut Eller -
Tobias C. Rittweiler