GPL code in Unixlib

Nick Burrett nick at sqrt.co.uk
Sun Dec 26 08:53:05 PST 2004


Peter Naulls said:
> My understading of GPL issues is imperfect, but this doesn't seem
> correct.  After all, glibc contains a great deal of non LGPL licenced
> code.  Certainly if it's GPL, then this extends to the program, which
> is less useful. These explanations may be helpful in why we should be
> using LGPL:
>
> http://www.fsf.org/licenses/gpl-faq.html#TOCWhySomeGPLAndNotLGPL
> http://www.fsf.org/licenses/why-not-lgpl.html
>
> In particular, the point was to why glibc is LGPL.

OK, that makes perfect sense and the Lesser GPL is the 'correct' license
for GPL files that we use.  Linking with files in a library that are GPL
does require that the application become GPL, unless those files have the
special exception, such as libgcc.

>
>> > but it's possible that having GPL code in UL
>> > has been a sufficient barrier against people using what has now become
>> a
>> > very powerful C library for RISC OS.
>>
>> Well it makes me wonder.  I though the barrier was code size :)  Even
>> though sticking a 500Kb module into memory so that we can have smaller
>> UnixLib binaries would probably cause a greater memory hog.
>
> UL typically adds around 100KB to a small program.  However, this is
> often dominated by other much larger libraires.  GTK2 for example, with
> its proliferation of requirements, means a 4MB binary minimun.  This is
> why I think we'd much rather have dynamic linking.
>
>> It may be possible to extract the code from lib1funcs.asm in GCC
>> instead.
>> I think it is GPL with special exception.  Mind you, isn't there code in
>> lib1aof.s that performs this job ?  That is certainly GPL with special
>> exception.
>
> I understand that Norcroft generates references to functions in ways
> that are not required by GCC, hence the requirement for this code.

Yeah, but the only difference will be the function name and calling
convention, which would not be a significant enough change to the original
code.

>> No need to worry here.  I clarified the issue with Huw Rogers a few
>> years
>> ago -- he was quite surprised that UnixLib still existed.  Peter and I
>> were in agreement on copyright issues some years back.
>
> I have a pretty good idea what this stance is, but perhaps you can
> repeat it once more for everyone's benefit.  Once this is resolved, I
> would like to make a clear statement on the GCCSDK site regarding
> copyright and licences of various items.

Huw Rodgers declared "no interest".
Pete Burwood's interest was the same as mine, which is "no interest".
Simon Callan had a different opinion at the time, but in more recent
years, declared no interest.

If we're going to do this properly, then we should really elect a
copyright agreement to follow and then ensure that any contributors sign a
past-and-future changes assignment to ensure that we can't have our asses
sued in the future.

The theory behind the Linux copyright model, in which nobody actually
needs to sign a copyright assignment, was on the premise that if enough
people wrote and changed the code, then there wouldn't be much to claim
copyright on.  The result being the lawsuits brought by SCO et al.

Overly paranoid ? Perhaps.  Once it is done, it's done though.


Nick.





More information about the gcc mailing list