Re: Targetting Sparc/Solaris

[Nathan Kidd <nathan-svn at spicycrypto dot ca>]

Sam Brightman wrote:
> Nathan Kidd wrote:
>>
>> If you have no sparc box at all, then in my opinion, the only sane
>> approach is to go to ebay, pick up an older sparc box with solaris 9 for
>> a few hundred bucks and then you can a) get your native headers/libs and
>> b) actually be able to test!
> 
> Are we/you then not back with the original problem of depending on the
> old Sparc box in several years time?

I don't think so.  From a building perspective, you need the box for
*initial* compiler setup, but once that's done you've got all the
benefits I mentioned earlier -- you don't need the box for *building*
any more.  Your QA process is free to do whatever (with a local SPARC
box or otherwise). If the box dies nobody really cares (e.g. QA only
needs a generic SPARC box, not a particular machine with specific disk
space, mounts, compiler versions, etc.).

> As a final attempt - is there any reason why a statically compiled
> version of our app built with crosstools targetting Sparc shouldn't
> work? All the code is built targetting Sparc, and surely and glibc etc.
> functionality is in the binary?

If you're making a native Solaris 9 binary then glibc is not part of the
picture; you should link with Solaris 9's libc. I'm not an expert on how
the symbols are resolved, but if your goal is to produce a binary that
will "just work" on any Solaris 9+ then I don't think there's any way
you can get around needing the native libs.

Happy building,

-Nathan

--
For unsubscribe information see http://sourceware.org/lists.html#faq