Re: FastCGI / Apache problems

Mark Brown (mbrown@OpenMarket.com)
Mon, 26 Aug 1996 16:42:18 -0400

Message-Id: <199608262042.QAA19429@breckenridge.openmarket.com>
To: fastcgi-developers@OpenMarket.com
Subject: Re: FastCGI / Apache problems 
In-Reply-To: <199608262000.QAA11663@bill-graham.nfic.com> 
Date: Mon, 26 Aug 1996 16:42:18 -0400
From: Mark Brown <mbrown@OpenMarket.com>


Bob Ramstad says

    i've also recently seen some signal 13, SIGPIPE, errors which also
    cause the FastCGI process to die and be restarted.

This reminds me of a couple of other bugs I noticed as I was in
there working.  Who knows, maybe you are tickling one of these:

 a) mod_fastcgi sends the "keep connection" bit; it shouldn't.
    The fix is in the function SendBeginRequest, which should
    call MakeBeginRequestBody with second argument FALSE
    rather than TRUE.

 b) mod_fastcgi does not correctly send and end-of-stream
    indication on the stdin stream.  If your application
    tries to read more than CONTENT_LENGTH bytes from
    stdin, it could hang.

Both SIGPIPE and SIGSEGV should be giving you core dumps
which will allow debugging to proceed.  I just forced
a core dump of a fastcgi application running under
mod_fastcgi on a Solaris 2.5 system and the core file landed in the
directory containing the executable.  This is the working
directory of the application as set by OS_ExecFcgiProgram
in mod_fastcgi.

    --mark