Subject: Re: [ublas] [bindings] New traits system
From: Rutger ter Borg (rutger_at_[hidden])
Date: 2010-01-14 09:32:44
Thomas Klimpel wrote:
>> Do someone know of C++ libraries
>> using this for some of their containers?
> Good question.
Well, in that case, let's start supporting it until someone really needs it?
I've fixed the overload resolution in case of a lapack binding with
workspace, with differentiated free functions for the real and complex
cases: in that case it could be that two functions existed with the same
number of arguments, one with Workspace as last argument, and one with, e.g.
MatrixVS as last argument. I brought in enable_if/disable_if to fix it. This
fixes the regressions for geev and gees.
The only remaining failing regressions are those for gbsv and hbevx. I think
the banded stuff is rather exceptional (the storage signature is changed by
the function), let's give it some more thought.