Re: Why not gnat Ada in gcc?

[Mike Stump <mrs at windriver dot com>]

> To: Mike Stump <mrs@windriver.com>
> Date: Wed, 01 Nov 2000 22:04:31 -0700
> From: Jeffrey A Law <law@cygnus.com>

> Exactly.  In fact, Geoff's automatic regression tester and "email
> the offending parties" is proving quite useful in identifying
> culprits and getting them to fix their problems.

Cool!  I took another look at it, and wow! neat!  Glad to see some CPU
cycles spent like this.  Oh, and after reviewing most of the bitchmail
since July 3rd, 2000, it seems that the qulaity of gcc as measured by
the tester has remained at 0 + epsilon, were epsilon has ranged
between 0 and 11 or so.  So, there you have it, the quality is stable
and good.  The only thing left, is to get everyone running something
like it for all their real targets and systems they are interested in,
and ramp up the data.  I'll make a prediction, the numbers will remain
at 0 + epsilon, for all testing that we do.  If you want performance
to remain good, merely put it in the testsuite.

> Yes, Dewar, if someone breaks the tree, they get a nasty gram about
> it.  In fact, the bulk changes from ACT have been one of the biggest
> offenders in the past.

We need the system to tally up the hate mail stats, and present a
bozos web page of people that like to break and not fix the compiler,
followed by the people that break it and leave it broken the longest.

:-) It is amazing the coercive power I think we can find in such a
system.  Also, systems like this should measure the good side, and
remark when people do good.  (His does.)

So, I have disk and CPU to burn, is there is distribution that has
enough bits to let me pump more data into the process?  I currently
pump only via test summary.  A good start.  If the answer is no, I'd
like to request people consider putting the source in contrib in cvs.