https://bz.apache.org/bugzilla/show_bug.cgi?id=61222

--- Comment #4 from Ruediger Pluem <[email protected]> ---
(In reply to Joe Orton from comment #3)
> Created attachment 35316 [details]
> rip out buffering in content length filter
> 
> I feel like this code should work harder to justify its existence, or else
> we should rip it out.
> 
> If the server should buffer some output specifically to try to compute a C-L
> and avoid chunked output for streamy content generators (e.g. CGI), OK,
> there's surely an argument for doing that.  But clearly you need to think
> hard about limits, which was never done here and somehow we got away with it.
> 
> Trying to push buffering into arbitrary filters is annoying and probably
> wrong unless we are carefully looking at performance trade-offs.
> 
> Anyway, alternative is to rip out the buffering here, and fixes the bug.

Understand that you are upset and I am quite surprised how old this code is and
that we never got hit by it. Just a comment to the patch: It breaks the stats
of r->bytes_sent.
How about passing everything down the chain once we did an apr_bucket_read no
matter if  it was APR_SUCCESS or or EAGAIN?

-- 
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to