testsuite removal

Joe Buck jbuck@Synopsys.COM
Sun Jan 31 23:58:00 GMT 1999


[ I'm probably saying too much below ... please don't take any of this
  as final, it is still in flux. ]

Jeff Law writes:
> > > > The existing testsuite will have to be removed from the distributions.

I wrote:
> > We aren't talking about dumping the existing tests

Mike writes:

> Ok, so I am confused.  We are talking about removing the testsuite
> from the distribution, but not about dumping existing tests?!

Jeff and I aren't really disagreeing.  We're talking about removing
the existing testsuite from the distributions, but not "dumping" them;
they might be distributed separately with some kind of "unofficial"
status.  (Unless we manage to clear the copyright on almost everything,
then it may not pay to have a small separate test suite).

> still don't get it.  I can randomly guess at what both of you mean,
> but would be nice to see someone else say it more completely.

We can't be complete here because the exact final decisions haven't been
made.  All we can say for sure is that there are legal problems with the
testsuite and we are trying to clean them up.

> > It's unclear that the testsuite is even legal,
> 
> Is it then also unclear if we can have a message archive of egcs-bugs?

Dunno.  There are some protections for ISPs we might hide behind.

> > The issue isn't assignment (that is, we aren't saying that assignment is
> > required).
> 
> Ah, ok, I see, we are not going to ask for assignment (good), but how
> then do we put a FSF copyright notice on it then?

I said "we aren't saying" meaning that it hasn't been decided.  I don't
know what the requirement will be.  FSF could copyright the collection
even if some contained pieces are copyright some other way.  Or small
tests can just be disclaimed, or used by FSF under fair use.  We're still
waiting on some legal advice for the details.

> Basically the questions I have are:
> 
>   What will the loss be in terms of existing testcases.

That question can be thought of in two ways:

* will we prevent people from accessing the current test suite?  No.
  However, some mirrors might not want to carry the old test suite once
  these concerns are widely known, and the FSF doesn't want to, which
  becomes a factor if we're going to reunite.

* what fraction of existing tests can go into the new test suite?
  We don't know the answer yet.

>   What additional requirements do we place on bug reporters in the future?

Anyone can report a bug ... we can't force them to do anything.  It will
not be uncommon for developers of proprietary software to report bugs with
snatches of proprietary code in the bug report, without clearance from
the boss.  When a gcc developer sees a report from the net, to be safe
s/he has to assume that this is what happened.

>   On bug fixers who want to add testcases from egcs-bugs that they fix?

We'll need more input on this one.

> Thanks for your time, answer at your leisure, as while this is
> important to me, it isn't pressing.

As an interim measure I'd just ask anyone checking in new tests to include
comments in the tests about where the test comes from, so it will be
easier later on if we do decide we need papers.





More information about the Gcc mailing list