|
|
Ublas : |
Subject: [ublas] Fwd: Bug in 1.59 in file boost/numeric/ublas/detail/concepts.hpp
From: dariomt_at_[hidden]
Date: 2015-12-01 03:50:44
Corss-posting to ublas list:
To me it seems that the regression was introduced in commit
https://github.com/uBLAS/ublas/commit/d6e77ccdf47d8f9f7935f576b820cdb25263b95a
Regards!
---------- Forwarded message ----------
From: <dariomt_at_[hidden]>
Date: 2015-11-30 15:24 GMT+01:00
Subject: [Ublas] Bug in 1.59 in file boost/numeric/ublas/detail/concepts.hpp
To: boost-users_at_[hidden]
In the migration from Boost 1.55.0 to 1.59.0 I found a bug in Ublas.
For instance, in 1.55.0 function OneElement() declared in
boost/numeric/ublas/detail/concepts.hpp line 716 looks like this:
template<> double OneElement (double) { return 1.; }
However, in 1.59.0 I think there is a copy'n'paste typo, because the
corresponding function (which has been refactored) looks like this:
template<class T> T OneElement (T) { return T(0.0); }
Please note that the latter version returns a *zero*, instead of a *one*.
This same bug affects the vector overload of the function, but the matrix
version looks correct (it returns identity_matrix).
This looks like a copy'n'paste typo, from a similar function
called ZeroElement(), that, of course, returns a zero.
I'm not sure if there is time to get a fix for this in the 1.60 release...
Regards!