httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chuck Murcko <ch...@Topsail.ORG>
Subject Re: bug with proxy TRACE? (fwd)
Date Fri, 21 Feb 1997 16:55:43 GMT
> Chuck Murcko wrote:
> > 
> > Sigh. Apache isn't an HTTP/1.1 proxy yet. That's on the punch list for
> > post-1.2 release, probably 1.2.1. The proxy also needs major cache
> > performance improvements, which will follow with or just after HTTP/1.1
> > compliance.
> > 
> > I've had my hands full just getting the proxy to the point where it works.
> 
> I've got some bad news for you, then! I'm tracking down some problems for a
> client, and they come down to the proxy using note_cleanup_for_fd() and then
> calling bpushfd() which does the same. This means that when the pool is
> destroyed files get closed which shouldn't!
> 
> The simple fix is to remove the note_cleanup_for_fd() from bpushfd() - it isn't
> so easy to do in the proxy module itself, as quite often Other Stuff (tm) goes
> on between the note_cleanup_for_fd() and the bpushfd().
> 
> Or, we could make it OK to note the same fd twice, or we could have a
> note_cleanup_for_fd_if_not_already_noted().
> 
> Thoughts?
> 
Double sigh. I've already +1'ed this fix. It seems that simple ORish
behavior can transparently do this, but within bpushfd it would be necessary
to maintain a list of descriptors to check on each invocation, yes? Or would
the list need to go into each buffer?

chuck
Chuck Murcko            The Topsail Group             West Chester PA USA
chuck@topsail.org

Mime
View raw message