Several months ago, I installed a particular version of smlnj (110.5-something) from an rpm. It worked fine. The guy using it wants an older version now, so I uninstalled with:

rpm -e smlnj

And "rpm -q smlnj" verifies that nothing is installed.

Then I tried to reinstall using "rpm -ivh smlnj-110.0.7-411.i586.rpm" (AMD 64 chip, Suse Linux 10 OS) and upon running sml, sml says:

sml: unable to determine architecture/operating system

I've also tried installing several other versions (

Why does the uninstall use -e (wouldn't -u be more appropriate)? Is it possible that I uninstalled incorrectly and something is conflicting with the new installations?

Any other ideas?

"-e" (erase) makes sense too, as does "-u" (uninstall).

And if you look at what you've written, the problem is right there:

You're trying to install an x86 (in this case, i586), package, on the x64 platform. You need the right architecture to make it work. You can either download the AMD64 .RPM file, or you can download the source and compile it from there.

Aren't the newer 64 bit intel and amd chips backwards compatible with old x_86 programs?

I ended up installing the latest code from source successfully, but the guy still needs version 110.0

should i586 rpms work?

ok, I found the source for the version I need in an rpm. Why is the source in an rpm instead of a tgz, I can't say.

I also can't say where the source ended up after I "installed" the source rpm. It's not in the usual place ( /usr/share/smlnj ).

The src installed itself in the rpmbuild root dir.
It should be in either:
/usr/src/redhat (my case)
or /root/rpmbuild
or /home/<username>/rpmbuild

I am not familiar with the app but it will create a file called something.spec in the SPEC dir which in in the rpmbuild root path

go to this base dir and then type:
rpmbuild -bb ./SPEC/<filename>.spec

It will build the rpm needed for your system in the /<base dir>/RPM/<arch>

Also please note that if this is a regular application there is absolutely no reason to build packages as root account user.

well, I want the program installed for use by all users. it's a programming environment (ml)

Storing source files in an RPM packages can be an advantage if the distribution in question has a specific default location for storing source files. This concept made more sense in those days when disk space was limited (I still have a few 280 MB hard drives around) - nowadays it doesn't even matter if you have the same sources stored a few times in different location, as they take less space then the audio and video file collections.

Back to the topic:
You should be able to run 32-bit binaries on a 64-bit system, but this is risky in some cases where the application does bitwise operations. In such a case a 'long' (usually 32-bit bit, but can also default to the native CPU bit-width) and 'long long' (should be 64-bit always, until 128-bit CPU's become de facto standards) can yield different results. The application will not crash, it's worse: you get trash results.
So be careful with such an approach. Also note that the above problem is not resolved by simply recompiling under a 64-bit system: the source need to be adjusted.

Arch Linux

but...

athlon64's were around for a very long time and caused no problems for windows users that I know of (because, I thought, they were completely backwards compatible).

I mean, I have a B.S. in Computer Science so I totally understand where you're coming from here, but I thought the CPU designers made it backwards compatible at least for already compiled 32 bit programs. EDIT: perhaps the difference is that my OS is 64 bit in this case, not just the CPU?

I just tried compiling this program from source and it fails because it cannot detect the architecture (hijacking the detection script and forcing it to be x86 causes it to fail spectacularly as it tries to compile instead).

Fun stuff.