Henrik Nordstrom wrote:
On tis, 2007-11-27 at 23:27 +1300, Amos Jeffries wrote:
Sigh. We need to clean up these MD5 issues before they propagate further.
Not sure how much issues Squid-2 is having with it at the moment.. I am
fine with leaving it as it is until people complain.
I'm just thinking at lest one of Adrians testers immediately brought up
the compile clash that forced --enable-openssl to be used as workaround.
Thats a little two-faced IMO. Quoting a GPL clash as to why we don't
build with SSL, then requiring it built in by our users in order to work.
AFAICT the squid-3 naming change was designed to fix that. Though
whether it also fixes the underlying problem of linking the right code
at the right time ... that was my Q below.
Has the renaming fix in Squid-3 shown ANY problems since Duane committed
it? If not, I'm proposing a cross-over to make the Squid-2 code identical.
Had to fix the OpenSSL glue (in both versions). Working compile with
--with-openssl is a MUST. If it can use other MD5s then that's fine..
Note: The bundled MD5 implementation is a quite poor performer.
Yeah.
Regards
Henrik
Amos
