Resolved: [Patches] options for filedup.

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

At 04:49 PM 3/21/2003, William A. Rowe, Jr. wrote:

>apr_dupfix.patch keeps the cleanup_kill -> dup2 -> register_cleanup
>logic that protects us from dup2'ing into an apr_file_close()ed fd.
>
>apr_duprevert.patch keeps the reorganization of the code, but drops
>the cleanup_kill and register_cleanup in the dup2 case, in favor of 
>trusting the current cleanups (and that they correpond to ->flags.)

I commited the second apr_duprevert flavor that reflects Joe's concern 
with too many recent changes.  This restores the old behavior of _dup2().

>In *BOTH* patches we revert apr_file_dup() assumption that fd 0..2 are 
>inherited, to reflect Jeff's concern with that change.

This patch was applied separately, as I agree with Jeff and expressed
the concern in my original commit message.

I've checked out on OS/X... others are encourgaged to cvs up their 
apr trees and see if the oddities introduced Wed night are sufficiently
resolved.  I have gone ahead and tagged the final tag and will roll in
the morning if Sander or Aaron don't beat me to it.  (I need to do
some magic to get a good build schema with all the bits that the
release.sh script expects.)

Sorry for the temporary breakage, and my enthusiasm for addressing
these real sorts of concerns with only Joe, and later Jeff also willing
to participate in an effort in a usable and yet somewhat more secure
release.

Bill