Marinier, M. Claude, G. wrote:

I am trying to port a 3D graphics package from VMS/Alpha with DEC Pascal to Solaris 7 (a.k.a. Solaris 2.7). I have installed version 19990119 of gpc with version 2.8.1 of gcc.

I have successfully compiled the first three levels of the graphics package. When compiling level4.pas, I get a 'fatal error 6' message, a .gpi file but no .o file.

I systematically trimmed code (whole functions and procedures) off the end of the source file until is compiled. Most of the implementation section was gone at that point. I tried commenting out the same code with '(*' and '*)' with the same result: it compiled. I then tried to remove only the code on the border, i.e. the compilation failed, I removed another procedure, and it compiled; that last procedure is the one I removed from the full source. This was based on the assumption that some code triggered the failure. The compilation failed with the 'offending' code removed. I abandonned that avenue of research.

The bug might be provoked by a combination of two procedures, so this effort wasn't in vain. If the shortest code that still fails is not very long (I suppose it isn't if most of the implementation was removed -- and probably much of the interface can be removed as well), you might want to show it to us so we can have a look at it and try to fix the problem.

Try using the latest gcc as a base?

Might help. Also, a new GPC snapshort will be released soon (which will require gcc-2.95.x), and you might want to try it then.

Use Modules instead of Units? I would need documentation of those.

Probably not. Units generally work better than (or at least as well as) modules in GPC.

Frank