> The connection to upstream (e.g. httpd) is closed so the client gets a 500 
> error.

Hmm, this isn't my experience. Possibly a slowcgi bug? My clients were
getting no response, e.g.:
curl: (52) Empty reply from server

> But the script keeps running.

Ah, I see now that the process is left to run now. My CGI read/writes
several MB/s lasting longer than the hard coded 120 sec timeout which
means my process would immediately get a broken pipe which led me to
erroneously believe my process was being terminated by slowcgi.

> We were worried that scripts are not prepared when they suddenly get 
> terminated.

To put worry at rest, in the rfc 3875 CGI Version 1.1 section 3.4 Execution:
...
"In the event of an error condition, the server can interrupt or
   terminate script execution at any time and without warning.  That
   could occur, for example, in the event of a transport failure between
   the server and the client; so the script SHOULD be prepared to handle
   abnormal termination."

> Also as sthen pointed out your MUA mangled the diff.

I believe I mangled the diff by copying from my terminal window.

> the man page needs better wording.
> please use strtonum(3) and use the idiom from the man page example.

I will attach the patch.

-alfred

Attachment: slowcgi-t.patch
Description: Binary data

Reply via email to