Matt Austern dixit:
I suppose it's time to ask: *should* Pascal integration be a goal for 4.1?
What about the goal being: Pascal should at least compile and work for simple programmes for 4.1 and be "delivered with", but not built by default, and full integration being the goal for 4.2 (if it's made in time for 4.1, even better)? In my experience, lower goals tend to be more easily achieved.
I'll shut up from now and leave it to the GPC maintainers to express what they want (I think Frank has already said that).
//mirabile
Thorsten Glaser wrote:
What about the goal being: Pascal should at least compile and work for simple programmes for 4.1 and be "delivered with", but not built by default, and full integration being the goal for 4.2 (if it's made in time for 4.1, even better)? In my experience, lower goals tend to be more easily achieved.
Actually, that's quite a high goal. To repeat myself, the real work is not in moving some files or doing some administrative work, but actually porting the code to the backend changes. Most of these changes will be needed even for simple programs. (Of course, there are parts of the compiler that a hello world program will not exercise, but that would be a strange division, and for any useful program, you could probably leave out a small fraction of total required changes only.)
Thorsten Glaser wrote:
Gabriel Dos Reis dixit:
The situation for 3.3.x is radically different from that of 3.2.x.
The files in gcc/gcc/config/ from 3.3 are radically different from 3.2, which made it a) not feasible for MirOS to upgrade to 3.3 b) not easy for gpc (if I remember the mailing list stuff correctly).
Yes, what affects us are the different target conditionals. (At least we have a partial workaround, and plans how to fix the issue -- which is actually a bit more general, and we've had a different form of it on MIPS/IRIX for a long time, so we'd have had to fix it sometime, anyway.)
Frank
-
Frank Heckenbach -
Thorsten Glaser