From: Sam Joseph <sam@...>
Date: 2006-02-08T23:16:05+09:00
Subject: Re: odd http header corruption

Hi Zev,

Zev Blut wrote:

> This looks like you have a "puts someobj.inspect" in your code.
> Depending upon how you are running your system this rogue puts is
> being called before the header output and is going directly to the
> output stream.  This is why your attempt to cut it out does not work.

Indeed I have suspected as much.

> I would grep your code for an "inspect" and if that fails "puts".

I had already done so - I have no "puts" or "print" outside my test 
classes, and no inspects outside of my logging code.  I've used the 
eclipse search function to scan all the code, and find | xargs grep on 
the server.  The only things that I can imagine being remotely 
responsible are things like:

./app/models/message.rb:      raise "more than one read_message for this 
user/message pair #{self.inspect}, #{user.inspect}"

and

./app/controllers/application.rb:      logger.info 'error during 
workspace log: ' + e.inspect

but these wouldn't affect stdout right?

I'm sure there must be something dumping that information somewhere, but 
I can't find it in my code, or in rails code, and more disturbingly to 
me, I can't work out where in the rails code the first part of the http 
header is generated, i.e. 'HTTP/1.x 200 OK', since the first call to 
output in cgi_process.rb is:

output.write(@cgi.header(@headers))

and @cgi.header(@headers) doesn't produce the first three lines of the 
http header as far as I can see, yet there are no other calls to output 
anything before that.

Anyhow, I've worked around the problem by printing a new line before the 
@cgi.header(@headers) is printed, which has fixed things for the time being.

CHEERS> SAM