From: Michael Widenius
Date: January 25 2001 4:01pm
Subject: building MySQL on OpenBSD with --enable-assembler
List-Archive: http://lists.mysql.com/internals/385
Message-Id: <14960.19941.769008.29015@narttu.mysql.fi>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Hi!

>>>>> "Tobias" == Tobias Oberstein <Tobias.Oberstein@stripped> writes:

Tobias> This applies to MySQL 3.23.32 + BerkeleyDB 3.2.3h.
Tobias> The --enable-assembler options of configure allows for
Tobias> compiling in x86 (and sparc) versions of common string
Tobias> operations, which should result in more performance.

Tobias> Apparently, the 2 involved files (for x86) strings-x86.s and
Tobias> longlong2str-x86.s won't compile on OpenBSD 2.8 / i686 / gcc 2.95.3.

Tobias> I've patched these 2, so they will compile:

Tobias> 1.
Tobias> Removed C-style comments not starting on Pos 1 of line with
Tobias> Shell-style comments aka '#')

Tobias> 2.
Tobias> modified "clrl %eax" to "movl $0, %eax" and the like ..

Tobias> 3.
Tobias> added a period to a segment size specifier

Tobias> 4.
Tobias> modified ".p2align 4,,7" pseudo-op to ".align 4"

Tobias> NOTE. All tests from the _mysql-test_ dir run successfully (except
Tobias> the one's related to Master-Slave replication), that are 94 tests.

This could be a timing issue;  We have fixed one case like this for
3.23.33...

What error did you get?
Did you have any error messages in the log files under var?

Tobias> WARNING: As I'm _not_ a x86 assembler guru, I'm not 100% sure if these
Tobias> modifications are safe. Any comments are welcome.

Tobias> The patched versions are included.

Tobias> Greets, Toby.


Tobias> "Anyone who considers arithmetical methods of producing random digits
Tobias> is, of course, in a state of sin."   John von Neumann

The above comments looks ok.  Unfortunately our mail handler removed
our attachment;  Could you send it to me so that I can merge
this with our code?

Thanks!

Regards,
Monty