solaris fixes

[Werner Koch]

Brian Warner <warner at lothar.com> writes:
> 
>  checks/mds.test: the $(foo) syntax requires bash. Using `foo` works fine
>                   in this case because there are no really weird shell 
>                   expansions inside.

Done.  And I learned that Solaris does not have a POSIX shell.

>  checks/run-gpg: solaris fgrep cannot take patterns from stdin. I moved
>                  the patterns to a file "run-gpg.patterns" and -f that
>                  instead of using "-f -" and a here-doc.

Done.

>  cipher/tiger.c: the byte array in TIGER_CONTEXT is accessed as a 64 bit 
>                  quantity at some point. This crashes processors that require
>                  64 bit alignment on those accesses. Shuffling the members
>                  to allow buf[] to be aligned solves the problem.

Done but not tested (I think it worked on an Alpha with disabled
alignment emulation)

> timeout in checks/genkey1024.test is too fast for my machine by a factor of 5

Increased.

> come up with a portable way of dealing with the missing __muldi3. Maybe just
> throw in a dummy multiply (of the same type used in.. was it tiger.c? one of
> the extension modules.. u32 * u32?) in g10.c somewhere and hope it doesn't get

I have put:

    {  /* a kludge to pull in the __muldi3 for Solaris */
       volatile u32 a = (u32)md,
       volatile u32 b = 42;          
       volatile u64 c;
       c = a * b;
    }

into  cipher/md.c:md_start_debug()


Werner