On 1/25/2012 2:28 PM, Jim Jagielski wrote:
> Bill, you should know by now that development is done on dev@...
> private@ is "on-list" as well, by your definition.
Jim,
You asked for bugs. We gave you bugs. We tracked notes on bugs. /sigh
> The issue is that 2.4.x is being held up by an issue, which
> is being "discussed" not on dev@, and since it's not a security
> issue, that's not the way we work.
It was your thread, dude; asked and answered. /sigh
That bug revealed a more sinister bug, with discussion on dev. /sigh
--- Begin Message ---
On 1/20/2012 7:07 AM, Jim Jagielski wrote:
> Just a friendly reminder... If reports have been submitted
> to BUGZ, could you list them in this thread please?
https://issues.apache.org/bugzilla/show_bug.cgi?id=52476
Suspects identified, proceeding to judgement and execution when
I wrap up my work-work today.
--- End Message ---
--- Begin Message ---
The main loop in the core output filter (rewritten since 2.2) will try
to read the entire passed-in brigade into RAM for CGI/PIPE-like mutating
bucket types. :( :( We have trying to bash this kind of bug since 2.0.x
days, and now the *core output filter* itself is doing it, yegads.
The fix may be as simple as adding an appropriate "break" to the loop,
I'm not sure yet. It should be doing the non-block read, FLUSH & retry
thing as well, so that streaming from slow generators works properly.
prefork/mod_cgi, serve this as "nph-killme.pl" and wait for the
fireworks:
!/usr/bin/perl
print "HTTP/1.0 200 OK\r\n";
print "Transfer-Encoding: chunked\r\n";
print "\r\n";
print "5\r\nhello\r\n" while (1);
--- End Message ---
