We're presently using an *old* (DOS) Pascal compiler on SCO Unix. Unfortunately, we've reached the end of the usable life of this compiler. The replacement we've provisionally chosen is FPC.

I've been using the Windows version of the IDE and compiler to convert my source files, and am quite pleased with the results.

However -- Am I correct in my understanding that I need the FPC compiler to compile the FPC compiler? If so, how far back to I have to go version-wise to get a compiler that doesn't require the FPC compiler to compile? As no SCO install is available (and I understand why), I'm kinda between the rock and the hard place.

Any and all help is greatly appreciated.

This won't help you. The problem is that neither the compiler nor the runtime know about SCO Unix. If you'd a compiler and rtl supporting SCO unix, you could easily cross compile from windows or linux or whatever. This is also the way how new targets are added: first, the compiler gets support to target the new OS. This compiler is a cross compiler e.g. linux -> sco. Then the rtl gets support for the new OS. Using the cross compiler this rtl is compiled. Now the cross compiler and the cross compiled rtl can be used to build a native compiler.

Daaave wrote

However -- Am I correct in my understanding that I need the FPC compiler to compile the FPC compiler?

Yes.

No, you can perform so-called "cross compiling": compile from one OS/ CPU for another OS/CPU. See the following pages for more information on how to (cross-)build FPC:
* http://www.stack.nl/~marcov/buildfaq.pdf
* http://wiki.freepascal.org/Cross_compiling

Jonas

PS: note to others: David is not subscribed to the list, so CC replies to him

Op Mon, 5 Feb 2007, schreef Daaave:

Yes.

If so, how far back to I have to go version-wise to get a compiler that doesn't require the FPC compiler to compile?

1.0.10. You need Borland Pascal though.

Don't take this route though. The compiler can generate scripts to assemble and link on the target. So, add SCO support, make the compiler compile itself for SCO on Windows, copy the files to SCO, and run the assemble and link script.

As no SCO install is available (and I understand why), I'm kinda between the rock and the hard place.

Well, I don't see why one would bother with SCO nowadays. However, if you want to do it, be our guest :)

Daniël Mantione

Didn't Carl mess with SCO at some point during 1.0.x times?

But indeed, the problem is that with lesser used proprietary OSes porting is a problem because maindevelopers have no interest/access.

This ranges from Windows 64-bit via Solaris and Morphos to OS/2 and recent BeOSes etc, as the effectively unmaintained part of the scale

And then I'm not even pulling in the general anti SCO sentiment.

This ranges from Windows 64-bit

I use and maintain Win64 daily?

So does Peter with Solaris. And Tomas(/Daniel?) with OS/2. But it rather limited compared to the core platforms (even OS X, e.g. Jonas, Vincent and me all have macs)

There is a reason why they are in that order. Win64 and Solaris are the more maintained ones of the row.

Op Tue, 6 Feb 2007, schreef Marco van de Voort:

True. But I consider i.e. Solaris different than SCO. Staying with SCO is just waiting until the building collapses. The rational reasoning is to get out of the building.

This has been true for OS/2 too of course, but unlike SCO it had good hobby value and active user communities. Beos and MorphOS rely solely on hobby users, and therefore hobby porting makes good sense.

Daniël

It's not a matter of wanting to, it's a matter of having to. Business decision versus technical decision, and unfortunately, not mine to make. Our customer base is running on SCO versions ranging from 5.0.2 to 5.0.7. Some of them have been with us for 15 years or more.

I'm in the process of trying to talk management into giving some flavor of Linux a shot, but we aren't there yet.