Other *NIX apps

[Kenneth E. Lussier]

On Wed, 2003-04-16 at 14:03, Jason Stephenson wrote:
> Kenneth E. Lussier wrote:
> > Hi All,
> > 
> > I know that all *NIX versions (Solaris, HPUX, Linux, etc.) have their
> > own little nuances, but deep down (and in some cases, really, *REALLY*
> > deep down) they are supposedly all the same. This makes me wonder: Is it
> > possible to run an old Solaris-based program on a Linux system? Just how
> > interoperable are all of these different *NIX system? Just some fodder
> > for discussion....
> 
> 
> That depends.
> 
> Marc assumes you're talking about binary compatibility. It's possible 
> among certain architectures and OS environments, but not very commonly 
> used. FreeBSD'ers seem to use it the most to run those few Linux apps 
> that we can't get native versions of, and of course, there's WINE, but 
> that's a different question.
> 
> Supposedly, there is this IBCS standard that if you have two operating 
> systems on x86-architecture that conform to the standard, you can just 
> run the app on either OS. I'm not sure what configuration or library 
> mangling you'd have to do, or if it even works. I've only heard about 
> it. I've never seen it working in practice.
> 
> In the case of source code, most applications are easily switched from 
> one modern UNIX to another. However, there are some issues that always 
> come up.
> 
> I've run into issues with system calls not always working the same from 
> one OS to another. poll() and sendfile() are two that come to mind 
> because I've had to deal with them recently. As an example, I ran into 
> an issue when working on KMail where some code that worked perfectly 
> fine on a GNU/Linux machine would hang under certain circumstances when 
> run on FreeBSD. I traced it down to the poll() system call. It has (or 
> at least had) different return value semantics on FreeBSD and Linux.
> 
> Then, of course, there are compiler issues. HP-UX doesn't ship with a 
> "working" compiler. Well, it's an old-style K&R compiler. To get an 
> ANSI-compatible compiler on HP-UX, you have to pay extra. Header files 
> often differ, too. POSIX had done alot to aid software portability, but 
> it isn't a panacea.
> 
> Beyond that are library issues, such as they get put in different places 
> by different UNIX vendors, and some libraries that you might expect to 
> be there (at least a GNU/Linux user might expect them to be there), just 
> aren't.
> 
> Well, I guess you really weren't asking about source code portability 
> after all, but being a programmer, I can't help thinking about it.

I wasn't asking about source code portability specifically, or binary
portability, either. My questions and comments stemmed from a
conversation that I had this morning where I was asked if I could run an
application on Linux that was written some years ago for Solaris. That
got me to thinking about the portability issues among *NIX versions. I
suppose that if you have the source code, then it would be a lot easier,
since you can change a lot of the dependant factors (if you know what
you're doing). But, the case that I had this morning was more of a "can
Linux do this, or do we need to find another way". I told them that they
probably needed another way since 1) we don't have the code 2) they want
it to run on an x86 system and it was written for SPARC, and 3) I just
don't have the time to muck about with it. But, I recognized it as a
question that I really don't know the answer to, and I haven't seen it
come up before. In my mind, that would be a major question for anyone
that was thinking of dumping a proprietary *NIX for Linux.

C-Ya,
Kenny
 
-- 
----------------------------------------------------------------------------
"Tact is just *not* saying true stuff" -- Cordelia Chase

Kenneth E. Lussier
Sr. Systems Administrator
Zuken, USA
PGP KeyID CB254DD0 
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xCB254DD0