Keith M Wesolowski wrote:
> On Thu, Mar 15, 2007 at 11:10:32AM +0100, Roland Mainz wrote:
[snip] 
> > It depends on how you define "sufficient way to test". I was often
> > thinking about writing a OS/Net tree-wide patch for that... but I think
> > this would be wasted work as long as we have to match the "requirement"
> > that such a patch must be tested with all possible/imaginable codepaths
> > (which is impossible for a single engineer (and even more difficult
> > since I lack Sun's test suites) ... and I wish the barrier could be
> > lowered to "... must result in a booting and running OS/Net which can be
> > used for OS/Net development without problems... " - that would be a goal
> > which we could archive quickly) ... ;-(
> 
> I'm pretty sure we've been down this road before.
> That's not a high enough standard.

AFAIK the previous "standard" was AFAIK that we _must_ gurantee that we
have found&&eliminated all problems before we can do a putback - which
is impossible for obvious reasons (how should we find all bugs in such a
large codebase ? AFAIK not even the switch from Sun Studio 10 to Studio
11 had to match such a "quality standard" (uhm... and I could complain
that some things are still broken since that switch... ;-( )). AFAIK we
need a "high enougth standard" which we can reach - for example the same
level which is used when the compiler gets updated between minor
versions (for example Forte 6U1--->6U2).

> A full PIT run is closer but I'd really prefer to
> see an approach that looks for source or binaries with likely problems
> and focuses testing on those.  We were finding writes into string
> constants fairly late in the gcc effort and I have little confidence
> that we really got them all.

Neither do I have the confidence that the gcc work caught all problems.
But somehow we need a way to start this work, right ?

> > ... and as I said... the real problem is that Sun Workshop/Forte/Studio
> > defaults to writeable string literals - this issue needs to be addressed
> > ASAP...
> 
> If we had a time machine, that would obviously be the right thing to
> do.  Sort of like the 256-fd limit.  But you're welcome to suggest it
> to them; they make Major releases and often break stuff like this that
> we then have to go work around.

My idea was not about "breaking" things - I only want compiler
"defaults" which matches how the ANSI/ISO C standard and most of the
code was written. And I even suggested a backwards-compatibilty way
(e.g. "-xstrnoconst") to make sure that we don't screw-up customers like
gcc4.x did.

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) [EMAIL PROTECTED]
  \__\/\/__/  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
  /O /==\ O\  TEL +49 641 7950090
 (;O/ \/ \O;)
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code

Reply via email to