Re: [eigen] Compile bug: aligned_allocator, gcc 4.8.2

[Gabriel <gnuetzi@xxxxxxxxx>]

Just seen it, damm, thanks!

Think the problem is resolved.
Hg has tricked me, since I did not correctly understand what hg update does.

thought it checksout the HEAD (but i was on branch 3.2) which was wrong

thanks lot!!


On 05/18/2015 07:39 PM, Christoph Hertzberg wrote:
> Am 18.05.2015 um 19:33 schrieb Gabriel:
> > i pulled it from bitbucket
> >
> > and did hg update :-)
>
> Watching your original mail, it seems you are not actually using the 
> most recent version:
>
> > /usr/local/include/eigen3/Eigen/src/Core/util/Memory.h|729|note: void 
> > Eigen::aligned_allocator<T>::construct(Eigen::aligned_allocator<T>::pointer, 
> > const T&) [with T = std::__detail::_Hash_node<std::pair<const 
> > unsigned int, A>, false>; Eigen::aligned_allocator<T>::pointer = 
> > std::__detail::_Hash_node<std::pair<const unsigned int, A>, false>*]|
>
> There is no aligned_allocator::construct() method anymore (we inherit 
> from std::allocator, to get that).
>
> Christoph
>
>
> > On 05/18/2015 07:26 PM, Christoph Hertzberg wrote:
> > > It might be related to this bug:
> > > http://eigen.tuxfamily.org/bz/show_bug.cgi?id=503
> > >
> > > Where did you get your version of Eigen? Downloaded from our site and
> > > installed, or from some package repository?
> > >
> > >
> > > Am 18.05.2015 um 18:58 schrieb Gabriel:
> > > > The strange thing is, that on the cluster (which uses gcc 4.8.2 as
> > > > wellI)  it doesn't compile as well... (os is centos 6)
> > > >
> > > > I cannot remember exactly from which branch I switched. I guess it was
> > > > 3.2.2 (stable, no dev)
> > > >
> > > > Thanks a lot!
> > > >
> > > >
> > > >
> > > > On 05/18/2015 06:21 PM, Christoph Hertzberg wrote:
> > > > > Am 18.05.2015 um 17:54 schrieb Gabriel:
> > > > > > I recently switched to the newest dev version 3.2.90 and
> > > > > > the following compile bug (only under gcc 4.8.2, 4.9 compiles fine )
> > > > > > arised:
> > > > > > This code worked before or did I use something incorrectly up to 
> > > > > > now?
> > > > >
> > > > > I don't get any errors compiling your code (after adding #include
> > > > > <unordered_map>) using g++ (Ubuntu 4.8.2-19ubuntu1) and the current
> > > > > head of the dev-branch.
> > > > > I guess you either removed something essential from your code snippet
> > > > > to reproduce this, or you have a buggy c++-library ...
> > > > >
> > > > > From what Eigen version did you switch (i.e. what date or what
> > > > > hg-hash)?
> > > > >
> > > > > Christoph