openssh unix dev - Jan 2006 - OpenSSL compile flags

When building OpenSSL version 0.9.7i (and all prior versions that I've 
seen) on Linux, the compiler flags passed to 'gcc' include:

    "-O3 -fomit-frame-pointer -mcpu=pentium"

This occurs even when the output of `uname -m` and `arch` are "i486". 
I
have a few machines with hacked 'uname' and 'arch' commands that
will
report "i486" even when the real arch is "i686", for use in
compiling
binaries for other i486 systems.  Looking at the 'config' and 
'Configure' scripts, it appears that the arch value is guessed based on 
'uname', 'arch', and 'cat /proc/cpuinfo', in different
cases.  I guess
I'm running into a situation where the configure script is looking at 
"/proc/cpuinfo", which I can't do a quick-and-easy hack to give
false data.

Is there an option that I'm missing to pass a particular arch value to 
compile OpenSSL for?  Right now, I'm working around this by 'sed'ing
the
above-mentioned string out of the CCFLAGS variable in the Makefile and 
replacing it with "-O2 -march=i586 -pipe".  Is this not proper, or 
should it work OK?  (It seems to work fine, in running it, but there's a 
lot of things I wouldn't know about it.)

Also, as a side-note, most of the literature I've seen regarding
'gcc'
says that the "-mcpu=" flag is deprecated, and that
"-march=" should be
used instead.  Is there a reason why OpenSSL uses "-mcpu="?

Thanks for the time and help, here--OpenSSL and OpenSSH are wonderful 
tools, and I appreciate all of the effort that goes into them.

Thanks,
Ryan

Perhaps you meant to send this to one of the OpenSSL mailing lists _not_
OpenSSH?

On Tue Jan  3 08:19:27 2006, Ryan B. Lynch wrote:
> 
> When building OpenSSL version 0.9.7i (and all prior versions that I've 
> seen) on Linux, the compiler flags passed to 'gcc' include:
> 
>     "-O3 -fomit-frame-pointer -mcpu=pentium"
> 
> This occurs even when the output of `uname -m` and `arch` are
"i486".  I
> have a few machines with hacked 'uname' and 'arch' commands
that will
> report "i486" even when the real arch is "i686", for
use in compiling
> binaries for other i486 systems.  Looking at the 'config' and 
> 'Configure' scripts, it appears that the arch value is guessed
based on
> 'uname', 'arch', and 'cat /proc/cpuinfo', in
different cases.  I guess
> I'm running into a situation where the configure script is looking at 
> "/proc/cpuinfo", which I can't do a quick-and-easy hack to
give false data.
> 
> Is there an option that I'm missing to pass a particular arch value to 
> compile OpenSSL for?  Right now, I'm working around this by
'sed'ing the
> above-mentioned string out of the CCFLAGS variable in the Makefile and 
> replacing it with "-O2 -march=i586 -pipe".  Is this not proper,
or
> should it work OK?  (It seems to work fine, in running it, but there's
a
> lot of things I wouldn't know about it.)
> 
> Also, as a side-note, most of the literature I've seen regarding
'gcc'
> says that the "-mcpu=" flag is deprecated, and that
"-march=" should be
> used instead.  Is there a reason why OpenSSL uses "-mcpu="?
> 
> Thanks for the time and help, here--OpenSSL and OpenSSH are wonderful 
> tools, and I appreciate all of the effort that goes into them.
> 
> Thanks,
> Ryan
> 
> 
> _______________________________________________
> openssh-unix-dev mailing list
> openssh-unix-dev at mindrot.org
> http://www.mindrot.org/mailman/listinfo/openssh-unix-dev
> 

--
Iain Morgan