Re: [eigen] Re: PacketMath.h compiler error, GCC4.9 on Ubuntu 15.04

Thanks Nathan,

I have submitted bug 1085 as per your suggestion, although I feel a little sheepish about it as I'm not sure if the commit resolved an error that was uncovered and my lack of understanding is the reason I can't see that for what it is.

My main trouble is that I'm not quite sure if there is any issue within Eigen, or more likely if it is with my use case being incorrect. I was a little bit embarrassed to speak up until today because that is likely the case, although I don't understand what changed well enough to fix it. :}

When I make Eigen from the repository, everything is fine, and there are no errors within Eigen or the Eigen unit tests that I have run for the library itself. Additionally, the contents of the error messages seem to make logical sense in that there should be difficulty in assigning a long long int to an int, however, previously in the past year (until a few days ago), my code never threw this error, so, I'm not sure that there is anything wrong with Eigen, I just don't understand the change well enough to see how my usage is leading to the new error.

Any light you can shed on it for me is most appreciated!

Respectfully,

Mark Sauder

On Fri, Oct 9, 2015 at 8:03 PM, Nathan Yonkee <nathan.yonkee@xxxxxxxxx> wrote:

Thanks for tracking the commit down, I'll give it a look. In the meantime, you could post this to bugzilla ( http://eigen.tuxfamily.org/bz/ ) to see if anyone else has the same issue. It would be helpful to upload a file containing the terminal output so that we can see the full error text: the command to pipe the terminal output to a file and display it too is tee.

$ make check | tee upload_file

On Fri, 9 Oct 2015 11:44:46 -0600
Mark Sauder <mcsauder@xxxxxxxxx> wrote:

> Hello again,
>
> I have been able to definitively track the first occurrence of the compiler
> error to this commit:
> https://bitbucket.org/tbs1980/eigen-numerical-integration-module/commits/96cdf0c63790648a5e13289c03322c0de6d9ac33
>
> I'm not quite sure what to do about it yet, but will do a bit more digging.
>
> Thanks!
>
> Respectfully,
> Mark Sauder
>
> On Fri, Oct 9, 2015 at 11:01 AM, Mark Sauder <mcsauder@xxxxxxxxx> wrote:
>
> > Hi Everyone,
> >
> > I was curious if anyone else was seeing compiler errors like the following
> > in the past few days:
> >
> > /usr/include/eigen3/Eigen/src/Core/arch/SSE/PacketMath.h:584:42: error:
> > cannot convert ‘Eigen::internal::unpacket_traits<__vector(2) long long
> > int>::type {aka __vector(2) long long int}’ to ‘int’ in return
> > return pfirst(_mm_hadd_epi32(tmp0,tmp0));
> >
> > /usr/include/eigen3/Eigen/src/Core/arch/SSE/PacketMath.h: In function
> > ‘typename Eigen::internal::unpacket_traits<T>::type
> > Eigen::internal::predux(const Packet&) [with Packet =
> > Eigen::internal::eigen_packet_wrapper<__vector(2) long long int>; typename
> > Eigen::internal::unpacket_traits<T>::type = int]’:
> > /usr/include/eigen3/Eigen/src/Core/arch/SSE/PacketMath.h:584:42: error:
> > cannot convert ‘Eigen::internal::unpacket_traits<__vector(2) long long
> > int>::type {aka __vector(2) long long int}’ to ‘int’ in return
> > return pfirst(_mm_hadd_epi32(tmp0,tmp0));
> >
> > /usr/include/eigen3/Eigen/src/Core/arch/SSE/PacketMath.h:584:42:
> > required from here
> > /usr/include/eigen3/Eigen/src/Core/GenericPacketMath.h:299:13: error:
> > ‘typename Eigen::internal::unpacket_traits<T>::type
> > Eigen::internal::pfirst(const Packet&) [with Packet = __vector(2) long long
> > int; typename Eigen::internal::unpacket_traits<T>::type = __vector(2) long
> > long int]’ conflicts with a previous declaration
> > { return a; }
> >
> > It looks like it might track from this commit

> > <https://bitbucket.org/eigen/eigen/commits/14fc5626ff985262a01e8bb393322154ecdb0774>

> > but I need to do a little more testing to nail that down with certainty.
> >
> > Anyway, just wanted to ask if anyone was having that same trouble..
> >
> > Thanks for all of the work that is going on in Eigen right now, it
> > certainly is an amazing library. (Long live Eigen! :) )
> >
> > Respectfully,
> > Mark
> >
> > Sauder
> >

--
Nathan Yonkee <nathan.yonkee@xxxxxxxxx>