ANSI/ISO Compliance

ANSI/ISO Compliance

[Michael LeBlanc]

Has anyone used Plum Hall and/or Perennial to test ANSI/ISO
compliance of any reasonably recent version of GCC?  I'm
more interested in C++ than C.  Host and target don't
matter, although it would be really nice to see results
for a Solaris or Linux host and a PowerPC target.  Thanks.

Mike LeBlanc

Re: ANSI/ISO Compliance

[Geoff Keating]

Michael LeBlanc <leblanc@skycomputers.com> writes:

> Has anyone used Plum Hall and/or Perennial to test ANSI/ISO
> compliance of any reasonably recent version of GCC? 

Yes, several groups have.

> I'm
> more interested in C++ than C.  Host and target don't
> matter, although it would be really nice to see results
> for a Solaris or Linux host and a PowerPC target.  Thanks.

The Plum Hall licenses generally prohibit distributing such results.

-- 
- Geoffrey Keating <geoffk@geoffk.org>

Re: ANSI/ISO Compliance

[Alexandre Oliva]

On Feb 11, 2003, Michael LeBlanc <leblanc@skycomputers.com> wrote:

> Host and target don't matter

They definitely do, since GCC uses their C library.  The less
compliant it is, the more ISO C (and even C++) compliance problems
you'll find with the testsuites, and it won't be GCC's fault, since
it's not meant to mask off deficiencies of an OS's C library.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer

Re: ANSI/ISO Compliance

[Michael LeBlanc]

Of course they matter in general, but they don't matter in
particular, to me, right now.  So pick a Linux whose C/C++ libraries
are 99% GNU.  I would just like to see, without buying the suites,
how GCC/GLIBC performs.

Alexandre Oliva wrote:
> On Feb 11, 2003, Michael LeBlanc <leblanc@skycomputers.com> wrote:
> 
>>Host and target don't matter
> 
> They definitely do, since GCC uses their C library.  The less
> compliant it is, the more ISO C (and even C++) compliance problems
> you'll find with the testsuites, and it won't be GCC's fault, since
> it's not meant to mask off deficiencies of an OS's C library.

Re: ANSI/ISO Compliance

[Joe Buck]

On Mon, Feb 17, 2003 at 02:56:48PM -0500, Michael LeBlanc wrote:
> Of course they matter in general, but they don't matter in
> particular, to me, right now.  So pick a Linux whose C/C++ libraries
> are 99% GNU.  I would just like to see, without buying the suites,
> how GCC/GLIBC performs.

In the case of the Plum Hall suite, if you don't buy it, no one that
does buy it is allowed to tell you the answer.

Re: ANSI/ISO Compliance

[Richard Henderson]

On Tue, Feb 18, 2003 at 05:15:27PM -0500, Robert Dewar wrote:
> OK, so that actually is much more liberal than you had suggested, and
> means that if the FSF was a licensee, for sure it could state the level
> of compliance with the Plum Hall suite.

Somewhere I recall a clause in which one may publish only whether
the compiler fails or passes the testsuite as a whole, and *not*
the results of the individual tests.


r~

Re: ANSI/ISO Compliance

[Robert Dewar]

> 11.7 LICENSEE may include in its advertising or other documentation any
> true statement reflecting conformance of its own software to the unmodified
> and most current Suite++ LICENSED MATERIALS and Plum Hall hereby grants to
> LICENSEE a license to use the name "Plum Hall" in such advertising or other
> documentation, provided however that any such use shall be true and accurate
> and based on tests to verify compliance of its software with the unmodified
> and most current Suite++ LICENSED MATERIALS, and that LICENSEE shall provide,
> upon request of Plum Hall, detailed information documenting such statement
> or claim. Any such advertising or other documentation shall state
> that Suite++ is a trademark of Plum Hall.

OK, so that actually is much more liberal than you had suggested, and means that
if the FSF was a licensee, for sure it could state the level of compliance with
the Plum Hall suite. For anyone else to do it is subject to difficult interpretation.
The above is written with normal proprietary software in mind, so the phrase
"its own software" is not as clear as it might be :-)

One thing to remember about contracts is that they are written to protect interests.
There is nothing that stops you from breaking a contract, such an action merely
creates the possibility of claiming damages. But it is not damage *per se* to
break a contract, the injured party would have to show that there was actual
damage from the breach to claim damages. That's what I meant when I objected
to someone saying that the contract meant that something was not allowed :-)

Re: ANSI/ISO Compliance

[Michael LeBlanc]

Joe Buck wrote:
> On Tue, Feb 18, 2003 at 01:25:09PM -0500, Robert Dewar wrote:
> 
>>>In the case of the Plum Hall suite, if you don't buy it, no one that
>>>does buy it is allowed to tell you the answer.
>>
>>More accurately you sign a contract to that effect. Please don't create the
>>impression that breaking a contract is a crime :-)
> 
> This is exactly what I meant; nowhere did I use the word "crime".

Below is an excerpt from Plum Hall's current license.  My eyes glaze
over when I read stuff like this.  11.7 says to me that I can publish
results, but the burden of proof is mine.  If a customer buys my
compiler because of my claims of compliance, and he finds my claims
were somewhat inflated, he can sue me, but I can't turn around and
sue Plum Hall.  I think my interpretation is correct because I have
yet to find even one vendor who makes such a claim.

11.6 LICENSEE shall hold in confidence all results of using Suite++
LICENSED MATERIALS and shall not publish or provide such results to
others, without prior written approval of Plum Hall, except as provided
in paragraph 11.7.

11.7 LICENSEE may include in its advertising or other documentation any
true statement reflecting conformance of its own software to the unmodified
and most current Suite++ LICENSED MATERIALS and Plum Hall hereby grants to
LICENSEE a license to use the name "Plum Hall" in such advertising or other
documentation, provided however that any such use shall be true and accurate
and based on tests to verify compliance of its software with the unmodified
and most current Suite++ LICENSED MATERIALS, and that LICENSEE shall provide,
upon request of Plum Hall, detailed information documenting such statement
or claim. Any such advertising or other documentation shall state
that Suite++ is a trademark of Plum Hall.

Re: ANSI/ISO Compliance

[Joe Buck]

On Tue, Feb 18, 2003 at 01:25:09PM -0500, Robert Dewar wrote:
> > In the case of the Plum Hall suite, if you don't buy it, no one that
> > does buy it is allowed to tell you the answer.
> 
> More accurately you sign a contract to that effect. Please don't create the
> impression that breaking a contract is a crime :-)

This is exactly what I meant; nowhere did I use the word "crime".

Re: ANSI/ISO Compliance

[Robert Dewar]

> In the case of the Plum Hall suite, if you don't buy it, no one that
> does buy it is allowed to tell you the answer.

More accurately you sign a contract to that effect. Please don't create the
impression that breaking a contract is a crime :-)