Nikos Mavrogiannopoulos <[email protected]> writes:

> Another approach could be to put them into a version which
> has variable (per release name). That would allow linking with them,
> but will always break software that uses them.

I think I like this. Should be easy to implement, document and
understand. 

A version like

NETTLE_INTERNAL_@PACKAGE_VERSION@ 

in libnettle.map.in would do, I guess? By definition, this gets a new
value for every release, and we don't need to care about structure of the
version string (major, minor, patchlevel, rcX, whatever).

I guess I'd need to read up a bit on how linker scripts work.

>> Other symbols, e.g., _nettle_sha256_compress and _nettle_umac_nh, are
>> internal interfaces which might change if nettle gets new or
>> different optimizations.
>
> We can list the ones that are unlike to see incompatible changes and
> are useful in applications explicitly, and move any others to either
> local, or to a private version.

Right, we can decide case by case if those symbols should be in
installed headers (I think that's ok for the hash *_compress functions),
and if so, whether or not they should be document beyond header
comments.

Regards,
/Niels

-- 
Niels Möller. PGP-encrypted email is preferred. Keyid 368C6677.
Internet email is subject to wholesale government surveillance.
_______________________________________________
nettle-bugs mailing list
[email protected]
http://lists.lysator.liu.se/mailman/listinfo/nettle-bugs

Reply via email to