Re: New version of Bob's C compiler

[mjn3@codepoet.org (Manuel Novoa III)]

Hi Riley,

On Thu, Jul 25, 2002 at 08:56:25PM +0100, Riley Williams wrote:
> My original reply actually mis-read the question - I was referring to
> converting ELKS to require the latest BCC compiler, and my instincts say
> that we're being far too hasty at this point, and need to wait until the
> new BCC version is in general circulation before we do so. However, if
> the general concensus is that we should do so, I will fall in line.

Ah.  I too misunderstood.  I thought you were talking about mentioning
in the ELKS docs where to find the latest bcc.  I remember you had asked
me that very question several weeks ago.

> >> Who wants to conduct a poll to decide whether this is a good or bad
> >> idea? I can see viewpoints both ways...
> >> 
> >> FOR:     Much of ELKS becomes simpler with the latest facilities.
> 
> > I for one need the new features.  ;-)
> 
> I could make good use of them as well...but that is pretty much
> irrelevant to this discussion as we can already make use of the new
> facilities in our own projects whether we use them in ELKS or not.

True.  As I said above, I misunderstood.

> >> AGAINST: ELKS becomes dependent on having the latest BCC compiler.
> 
> > Is this really much of a drawback? Building bcc isn't that big
> > an issue.
> 
> Are you willing to be the one to deal with all the "ELKS doesn't
> compile" spam that will result from the people who download ELKS onto
> their system, then try to compile it and it throws up lots of error
> messages simply because they have an old version of BCC installed and
> we rely on a new version with lots of enhancements not in the commonly
> installed versions?

Not even _remotely_ likely.  ;-)  I see too much of that nonsense
already on the busybox and uClibc mailing lists.  I've got no
patience for it.

> I'm not - I had enough of that from HarKal when ELKS was sensitive to
> a difference between his system configuration and mine with the result
> that it compiled here but not on his system. We sorted that out a while
> back and removed the said sensitivity from ELKS, but this will bring in
> far more sensitivities than I'm willing to deal with.
> 
> As far as I can tell, RedHat issues a new release about every six weeks
> from what I've seen, and the other major distributors appear to issue
> releases at about the same rate. Also, apart from the major Linux
> developers (like myself), most Linux users I know keep their systems
> updated to within three releases of the current one.
> 
> My proposal would therefore be to work towards getting the BCC with
> these major enhancements included in all the major distributions, then
> give that four months to percolate through the userbase. At that point,
> 90%+ of the Linux userbase will have the relevant compiler, and we can
> start making changes that depend on those facilities without worrying
> about problems like this.

No matter what, there will need to be some test in the ELKS build that
waives a red flag for those trying to use an old version of bcc.  We
still see posts from people on the uClibc mailing list about bugs in
the older uC-libc that is being packaged with uClinux by some groups.
Just the other day, someone posted a message about printf behavior
that I fixed in uClibc over 18 months ago.  :-(

Manuel