Nikos Mavrogiannopoulos <[email protected]> writes:

> I've now have nettle-mini for both master and 2.7. The patches
> required for master are gmp-mini branch and for 2.7 in
> nettle-2.7-mini-gmp.

Nice!

> I'd appreciate if these go upstream as I've failed to add nettle to
> openwrt (and don't have to time to pursue that further), so I no
> longer plan to maintain these patches separately (and they will be
> easily break on makefile changes).

I'd really like to add mini-gmp support on the master branch, but I
think it has to wait until after the release. I see one issue which your
code, which I'd like to fix before integration, and that is that I think
it is a bit too brittle to link the test programs with the real gmp.
Problem is that mpz_t defined by gmp and mini-gmp is not really
compatible.

It's possible to use mini-gmp and gmp in the same executable, since the
symbol names seen by the linker are different, but then each compilation
unit should use either gmp or mini-gmp, and no mpz_t variables can be
passed across this boundary. The mini-gmp testsuite does this,
converting numbers to hex strings when they need to be passed between
gmp and mini-gmp functions.

For the nettle testsuite, I hope one can get by with something simpler.
Do as many of the tests as possible using mini-gmp only, and disable
tests which depend on functions only available in the real gmp library,
e.g., mpz_urandomb.

Regards,
/Niels

-- 
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
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