Allow solutions to write to their methods

[jgdumas]

For instance Wiedemann solver computes the rank of the matrix,
it would be nice to give that information back to the method.
For now all methods are const.

[bdsaunders]

I wholeheartedly agree. If we ever implement the postdoc probability reporting, it would also be written to the method. In examples/solve.C inconsistent systems throw a LinBox error. I would rather write that failure-to-solve reason to the method. In any case, if there is to be an error thrown, solve.C should catch it and give an intelligible message to the user. ...as part of this issue, have examples/solve.C report the rank as side information. Incidentally, Sano's little GF2 example fails intermittently for me in the blackbox case of solve.C. Stronger preconditioning still is needed.

…

On Thu, Sep 5, 2019 at 10:04 AM Jean-Guillaume Dumas < ***@***.***> wrote: For instance Wiedemann solver computes the rank of the matrix, it would be nice to give that information back to the method. For now all methods are const. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#232?email_source=notifications&email_token=AA4TUI6HRICCAXQPXEXXCXDQIEGW5A5CNFSM4IT6IS3KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HJRMX4A>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA4TUI3QAWVDAZEGPXL6QWDQIEGW5ANCNFSM4IT6IS3A> .