Re: [PATCH v6] checkpatch.pl: Add SPDX license tag check

[Joe Perches <joe@perches.com>]

On Fri, 2018-02-02 at 14:18 -0600, Kate Stewart wrote:
> On Fri, Feb 2, 2018 at 1:06 PM, Joe Perches <joe@perches.com> wrote:
> 
> > On Fri, 2018-02-02 at 12:27 -0600, Rob Herring wrote:
> > > On Fri, Feb 2, 2018 at 9:49 AM, Igor Stoppa <igor.stoppa@huawei.com>
> > 
> > wrote:
> > > > On 02/02/18 17:40, Rob Herring wrote:
> > > > > Add SPDX license tag check based on the rules defined in
> > > > 
> > > > Shouldn't it also check that the license is compatible?
> > > > 
> > > 
> > > Perhaps we shouldn't try to script legal advice.
> > 
> > True.
> > 
> > I believe what was meant was that the
> > entry was a valid SPDX License entry
> > that already exists as a specific file
> > in the LICENSES/ path.
> > 
> > So that entry must be some combination of:
> > 
> > $ git ls-files LICENSES/ | cut -f3- -d'/' | sort
> > BSD-2-Clause
> > BSD-3-Clause
> > BSD-3-Clause-Clear
> > GPL-1.0
> > GPL-2.0
> > LGPL-2.0
> > LGPL-2.1
> > Linux-syscall-note
> > MIT
> > MPL-1.1
> > 
> > From my perspective, it'd be better if the
> > various + uses had their own individual
> > license files in the LICENSES/ path.
> > 
> 
> At the end of december, the SPDX license list[1] was rev'd to
> Version: 3.0 28 December 2017.   At the request of
> FSF, the GNU license family would not use the "+" notation,

That's rather more sensible to me.

This should probably be updated in linux-next in
the near future rather than later.

> and would bias towards using "-only" and "-or-later", explicitly.
> So adding both variants to the LICENSES/ path aligns with
> this forward direction.

It's probably better to remove the + variants.

> > Right now, there are many missing licenses
> > that are already used by various existing
> > SPDX-License-Identifier: entries.
> > 
> > 
> > APACHE-2.0
> > BSD
> > CDDL

CDDL does not exist standalone and was caused by my
defective eyeballs when scanning the SPDX list via:

$ git grep -w "SPDX-License-Identifier:" | \
  cut -f3- -d":" | \
  sed -r -e 's/^\s+//' -e 's/\*\/\s*//' -e 's/\s+$//' | \
  sort | uniq -c | sort -rn

> > CDDL-1.0
> > ISC
> > GPL-1.0+
> > GPL-2.0+
> > LGPL-2.1+
> > OpenSSL
> > 
> > There are odd entries like:
> > 
> > GPL-2.0-only
> > 
> 
> This is the new way to represent GPLv2 only, as described above.
> While the GPL-2.0 and GPL-2.0+ notation is still valid,  it is deprecated
> in the latest version, so transitioning existing over time will probably
> be needed.

Probably better to remove and replace the old notation
instead of doing it piecemeal.

When the appropriate LICENSE file changes exist, a
generic substitution could work well.

$ git grep --name-only "SPDX-License-Identifier:" | \
  grep -vP "^(?:LICENSES/|Documentation/process/license-rules\.rst)" | \
  xargs perl -p -i -e 's/SPDX-License-Identifier:\s*(L?GPL-\d\.\d)\+/SPDX-License-Identifier: \1-or-later/;s/SPDX-License-Identifier:\s*(L?GPL-\d\.\d)(?!-or-later)/SPDX-License-Identifier: \1-only/'

> So I think the list of licenses to be added to
> LICENSES/ path is:
> 
> APACHE-2.0
> BSD
> CDDL
> CDDL-1.0
> ISC
> GPL-1.0-only
> GPL-1.0-or-later (note: actually same contents as one GPL-1.0-only)
> GPL-2.0-only
> GPL-2.0-or-later (same contents as GPL-2.0-only)
> LGPL-2.0-only
> LGPL-2.0-or-later (same contents as LGPL-2.0-only)
> LGPL-2.1-only
> LGPL-2.1-or-later (same contents as LGPL-2.1-only)

If LGPL-n.m -only and -or-later are the same,
there's probably no need for duplicate LICENSE
files and just making sure LGPL-n.m without any
other wording is exclusively used is proper.

> OpenSSL
> 
> Having files with the same contents, but different names is
> irritating, but I can't see a another way of complying with REUSE
> guidelines.   Any better suggestions?

What and where are the REUSE guidelines?

https://reuse.software/dev/
doesn't show me much.

> 
> 
> > Parentheses around AND/OR aren't consistent.
> > 
> 
> The SPDX specification has an appendix that calls for "(",")"
> around every license expresssion.   After discussion with some
> developers it was decided to be ok to relax that, and only add them
> when they were essential to clarify the logic.   The next rev of the
> SPDX specification will have this clarified as well.   I think we caught
> most of the changes in the kernel documentation patches for describing
> this,  but if you have specific cases to be reviewed,  happy to have
> a look.
> 
> Thanks, Kate
> 
> 
> [1] https://spdx.org/licenses/