RE: Generated source

["Kilbane, Stephen" <Stephen dot Kilbane at analog dot com>]

Thanks for the response, Ian. That's about what I expected,
and very useful.

steve


-----Original Message-----
From: Ian Lance Taylor [mailto:iant@google.com]
Sent: 29 March 2011 15:30
To: Kilbane, Stephen
Cc: binutils@sourceware.org
Subject: Re: Generated source

"Kilbane, Stephen" <Stephen.Kilbane@analog.com> writes:

> We're looking at adding contributing processor-specific code to binutils.
> The code in question would come from internal code-generation applications
> that would also contribute different but equivalent functionality to a
> proprietary toolchain. In other words, it's like having a yacc that squirts
> out two different y.tab.c files, one for binutils, and one for the
> proprietary toolchain.
>
> The question is: assuming all the licences and copyright-assignment
> paperwork's in place (it is), would such generated source be acceptable to
> the binutils maintainers, without also contributing the means of re-generating
> that source? I.e. without also contributing the source for the generation
> tool, and the definition files from which it works.

I think the binutils maintainers could choose to accept it, but I think
that technically this would not be acceptable under the GPL.  The GPL
requires that source code include "the preferred form of the work for
making modifications to it."  The case you are describing would be
source code, but not the preferred form.

Also problematic would be cases where the code needed to be changed for
portability reasons or due to changes in other parts of the binutils.
This would lead to differences between the code in the binutils and in
the master sources, complicating efforts to incorporate changes in the
master sources.

So all in all it's hard to see this as a good idea.  It would work once
but would cause increasing headaches over time.

Ian