Re: [squid-users] Solaris/OpenSSL/MD5 Issues

From: Amos Jeffries <squid3@dont-contact.us>
Date: Sun, 11 Nov 2007 16:32:38 +1300

Randall DuCharme wrote:
> Amos wrote:
>> Maybe this would be the better one if it works:
>> AC_CHECK_LIB(crypt, crypt, [CRYPTLIB=" -lcrypt "])
>> AC_CHECK_LIB(md5, MD5Init, [CRYPTLIB+=" -lmd5 "])
>
> This works fine. (still learning how to use these GNU auto-tools.
> I am a C/UNIX developer by profession but don't use GNU tools
> other than GCC occasionally. Thanks for the help and for your
> patience!)
>
>> If there were any other changes beyond the patch I sent and the
>> above could you send me a patch back and I'll queue it for any
>> future 26 releases.
>
> No changes other than your patch, copy/paste amendment to it, and
> the above configure.in addition. I noticed something though and I'm
> not certain of the desired behavior. If I use the --enable-ssl flag
> openssl is found and it appears to be enabled (looking at autoconf.h)
> Running ldd against the squid binary shows that it's linked against
> the openssl libs. The interesting thing is that I don't get any
> of those MD5 conflict messages like I thought I would. Without the
> --enable-ssl flag the Solaris MD5 header and library are used. If I
> provide the --with-openssl=, I can use an alternate OpenSSL
> installation ( SunStudio-built and aggressively optimized ) Is this
> the desired (new) behavior? Running configure --help says that
> --enable-ssl enables gatewaying support using OpenSSL so I'm assuming
> that it is. I'm just wanting to be sure that the OS's support
> is functioning correctly. What's the best way to thoroughly test
> this?

It is. OpenSSL is the currently preferred library. Any alteration in
that would require performance testing on all OS that provide their own
(just this Solaris10 so far, but maybe others later) to prove that its
better than the well-known OpenSSL code.

As for the testing. I'm not too sure myself. Performance and
benchmarking are largely outside my experience as a developer. I come at
things from an HCI/CHI and Architecture design approach.

>
>>> I'm not quite as lucky with 3.0.RC1 yet.
>
>> Okay. That looks like another seperate problem than MD5.
>> For some reason you have "-Wl,-R -Wl,/usr/sfw/lib/amd64"
>> twice times at the end of the g++ parameters, that looks to be
>> the problem. Are you manually setting those flags or leaving
>> to the autotools?
>
> I was setting -L and -R flags in in the LDFLAGS env variable during
> configure since with GCC I often have problems with the runtime
> linker not being able to find libraries and trying to use 32 bit
> libraries when doing a 64 bit build if I don't set them. I have
> removed all -L and -R paths (and I do not set LD_LIBRARY_PATH on
> Solaris) yet I still end up with the duplicate linker arguments.
> I'm off to see if I can figure out how this is happening.
>
> Kind regards
>

Amos
Received on Sat Nov 10 2007 - 20:32:40 MST

This archive was generated by hypermail pre-2.1.9 : Sat Dec 01 2007 - 12:00:05 MST