Re: portability as a goal for debian?
On Tue, 6 Mar 2001, Steve Greenland wrote:
>
> Some kinds of portability are relatively easy to achieve: many
> (most?) C language 64 bit cleanness issues have to do with people
> making assumptions about the size of "int" and its variations that
> are completely unwarranted.
This is very true. Porting software to architectures within an OS like
Linux is usually just a matter of making sure that the sizes of pointers
and ints (and all associated operations on such) are correct. The lesser
part is just adjusting to the quirks of each architecture that you're
dealing with.
> Other kinds of portability are difficult:
> making the same code base run on different OS's (not just different
> Unix variants, which have more or less the same system library, but
> completely different platforms, such as NT and OpenVMS) (Yes, I've done
> it. No, it's not pretty.)
I'll agree with this. I've done my share of porting software to different
OSs and it can be very arduous and painful. Even ports to a "similar" OS
can be ugly (try porting most Tru64 software to Alpha-Linux and you'll
understand, despite being on the same processor types).
> As far as the BSD port goes: Debian uses the GNU tools. Deal with it.
> Better yet, spend your time on something worthwhile, like the HURD port
> or the boot floppies.
I think that doing a BSD port may actually be a good thing, but
disagreeing with the "de facto" software that we use for Debian (like GNU
make) is something that should be in the "let sleeping dogs
lie" category. For us to change over to pmake or any of the other make
variants would be very very ugly compared to porting/using GNU make and at
least being that much further into the port. I use 'make' as an example
since it came up, but any tool can be substituted.
In all fairness, BSD makes good use of their tools and they work very well
for them. If a Debian BSD port is to really happen, that's just something
to work around for that port (insert creative solution here). This kind
of thing happens on a code level all of the time while I'm dealing with
the Alpha port. Most apps are written for 32-bit archs...that's just a
fact....and most of the time, if one of the 64-bit arch folks doesn't
write in the correct code or fix the software in some other way, it just
doesn't get ported. In short, we do quite a bit of work-arounds every
day. I don't see this as being any different really.
C
Reply to: