Is there any general guidance for moving headers out of boost/pending? Do we really need to include a forwarding stub for these? I'd like to consider headers in boost/pending as "not final" and therefore we should be able to move headers out of there without leaving a forwarding stub. Example: https://github.com/boostorg/dynamic_bitset/pull/24 - Jim
On 02/09/2018 03:03, James E. King III via Boost wrote:
Is there any general guidance for moving headers out of boost/pending? Do we really need to include a forwarding stub for these? I'd like to consider headers in boost/pending as "not final" and therefore we should be able to move headers out of there without leaving a forwarding stub.
pending contains "useful stuff that hasn't been reviewed" - in other words stuff that needed to be written to implement library X, but which doesn't form part of it's interface and wasn't reviewed and doesn't necessarily have the ultimate/best interface. So.... you would need to either: 1) Find another library where the component naturally belongs, and persuade them to adopt it (maybe with changes), or 2) Get it reviewed as a component in it's own right. Once it's "official" then the original library can be changed to use the new version.
Example: https://github.com/boostorg/dynamic_bitset/pull/24
- Jim
_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
--- This email has been checked for viruses by Avast antivirus software. https://www.avast.com/antivirus
On Sat, 1 Sep 2018, James E. King III via Boost wrote:
Is there any general guidance for moving headers out of boost/pending? Do we really need to include a forwarding stub for these? I'd like to consider headers in boost/pending as "not final" and therefore we should be able to move headers out of there without leaving a forwarding stub.
Go for a pragmatic approach? Look for those headers, say in github and stackoverflow. If you see many repositories where code includes the header, and many stackoverflow questions about them, try and avoid breaking them. If you don't see any, there may still be users somewhere, but they are likely quite few and you don't need to care as much. -- Marc Glisse
On 9/1/2018 10:03 PM, James E. King III via Boost wrote:
Is there any general guidance for moving headers out of boost/pending? Do we really need to include a forwarding stub for these? I'd like to consider headers in boost/pending as "not final" and therefore we should be able to move headers out of there without leaving a forwarding stub.
If the functionality in the headers is not documented you should be able to move the headers as you wish, as the headers are just implementation details.
- Jim
James E. King III wrote:
Is there any general guidance for moving headers out of boost/pending? Do we really need to include a forwarding stub for these? I'd like to consider headers in boost/pending as "not final" and therefore we should be able to move headers out of there without leaving a forwarding stub.
The general approach should be to leave the forwarding stub with BOOST_HEADER_DEPRECATED in it for a release or two, allowing existing users to migrate, then remove; but for this to work there needs to be an alternative to which people can migrate. Do we have an equivalent of lowest_bit in Boost.Integer? Is it just log2?
participants (5)
-
Edward Diener -
James E. King III -
John Maddock -
Marc Glisse -
Peter Dimov