Re: Best way of compiling applications to run on older linux distros

[Andrew Haley <aph at redhat dot com>]

Tom Quarendon wrote:
>> Can't you just link statically?  Then it's just a matter of being
>> syscall compatible.
> 
> Many issues with linking statically. Product is too big, and is
> architected as seperate shared libraries, so can't link libstdc++
> statically with each one.

Right.  In general, linking statically on Linux is a really bad idea.

>> Here's a tip, go grab yourself VMware or Sun's VM, then make a slew of
>> VMs and install the distros you want in those.  That way you don't
>> need an actual box for each OS and you can build it natively.  If what
>> you are doing is professional and not just some hobby project you
>> should just do it properly from the get-go.
> 
> The point is that I'm trying to work out what "properly" means. We do
> have a bunch of vmware boxes. That's how I know that it doesn't work
> when we build on our build machine and try and run it on Centos4.

So, let me see.  You have a Centos4 machine that you could use to build
(indeed, you need such a machine to do the testing) but for some reason you
don't want to build on that box as well.

> Your implication is that "properly" means building a version for Centos4
> on Centos4, a version for linux distro x on linux distro x etc etc. This
> isn't in any way obvious, and isn't how things work for Windows, or
> indeed Solaris, AIX, zSeries.

I've already explained backwards compatibility and how it works on Linux.

The question, Tom, is this: do you want help, or do you want to argue?

Andrew.