Re: Re: [eigen] Remove Eigen2 support for 3.2?

[ Thread Index | Date Index | More Archives ]

See, if the eigen2 support mode is fundamentally incompatible with the
evaluator changes, then _that_ is a really valid reason to drop it. By
all means, do not let anything stand in the way of the evaluator


2012/6/28 Gael Guennebaud <gael.guennebaud@xxxxxxxxx>:
> Alright, then we'll try to keep it around, however the evaluator
> changes might require us to keep only the stage 40. We'll see.
> Gael
> On Thu, Jun 28, 2012 at 4:24 PM, Marcus D. Hanwell
> <marcus.hanwell@xxxxxxxxxxx> wrote:
>> On Thu, Jun 28, 2012 at 10:04 AM, Laurent Rineau
>> (CGAL/GeometryFactory) <laurent.rineau@xxxxxxxx> wrote:
>>> Le jeudi 28 juin 2012 09:33:34 Benoit Jacob a écrit :
>>>> Is this a heavy maintainance burden or just a wish to simplify the codebase?
>>>> In the former case, go ahead and do whatever you need to keep Eigen
>>>> easy to maintain; in the latter case, I would suggest giving it a bit
>>>> more time and having a transition period where it would be officially
>>>> deprecated and complain loudly about it (generate warnings, or even
>>>> require people to #define
>>>> YES_I_KNOW_EIGEN2_SUPPORT_IS_GOING_AWAY_SOON), and only fully remove
>>>> it in the version after (3.3?)
>>>> This is just my 2 cents as an industry guy now ;-) People just don't
>>>> want their existing code to be broken, and value that much higher than
>>>> "cleanliness" and "removing cruft".
>>> I second Benoit's opinion. My company does not use Eigen2, but it is true that
>>> the industry does not like breaking changes for no real reasons.
>> I third Benoit's opinion, but would understand if it needed to be
>> removed. Having some mixed projects, and the ABI incompatible changes
>> in the Vector3d etc typedefs would perhaps make it harder for us, but
>> we could also require 3.0-3.1 if necessary. We are working on moving
>> everything over, but these things take time.
>> Marcus

Mail converted by MHonArc 2.6.19+