Re: Big vs. Little endian problem in md5 code?

[Jan-Henrik Haukeland]

Christian Hopp <address@hidden> writes:

> Hi!
> 
> I am implementing the md5 stuff right now for certs... so I "played" a
> bit with using the _file_ md5 support!  Result was this.  I took
> /bin/true as my target, like this...
> 
> check pop3s with pidfile /var/run/pop3.pid
>       port 995 type tcpssl protocol pop
>       start program = "/bin/true"
>       stop program = "/bin/true"
>       alert address@hidden
>       checksum /home/chopp/true expect 508b2fa3a4fe7139341a9a919660bab2
> 
> I have generated the md5 sum...
> 
> Solaris sparc with md5sum prg...: 508b2fa3a4fe7139341a9a919660bab2
> Solaris sparc with md5 prg...: 508b2fa3a4fe7139341a9a919660bab2
> Linux intel with md5sum prg...: 508b2fa3a4fe7139341a9a919660bab2
> 
> Monit Solaris sparc...: 8b2502d1c1c1acc2699c3f5c46afbf43
> Monit Linux sparc...:  8b2502d1c1c1acc2699c3f5c46afbf43
> BUT Monit Linux i386...: 508b2fa3a4fe7139341a9a919660bab2
> 
> And the file is of course always the SAME.
> 
> What's happening there???

Could it be the machine architecture? for instance that Solaris is 64
bit and Linux a 32 bit system and the md5sum algorithm in monit uses
the size of _int_. In that case it's strange since the md5sum source
we use in monit should be the same as the one the GNU md5sum program
is using, but maybe they have done some update of the code. I think
the code is in GNU binutils.

-- 
Jan-Henrik Haukeland