Just to follow up on the discussion, the main (and only, as far as I'm aware of) reason for using a gcc different from the one distributed with the developers tools is fortran support. This is an issue, as already pointed out, if you want to compile trilinos and petsc with reasonable features. I never got trilinos 10.0 to compile all the things I need without fortran support (which should be possible if you only need a small subset of trilinos features). Has anybody in the list managed to do that?
If so, I'd like to hear your comments on it (maybe a post on the wiki would make this topic less buried into the mailing list, since it looks like it comes up more and more often...). This would eliminate the need of an external gcc (that is, if you don't need petsc...). A second source of confusion is the fact that the gcc distribution which you find on gforge is not the same that you find directly from the HPC web page (at least that was the case the last time I tried). It looks like the one on gforge adds more problems than the ones it solves... I have not had casting issues, even with apple distributed gcc. This is another point where it is not clear what the real issue is. A particular combination of hardware and software? Triggered by additional/different libraries installed on newer/older macbooks/macbooks pro? I'd like to understand this a little more, and it could be useful if we could trace where the problem actually lies. If anyone has these casting issues, can they report gcc versions, os x version and hardware where this happens (i.e., MBP 2009 or MBP 2007)? Thanks in advance. Luca. -- Luca Heltai <[email protected]> http://people.sissa.it/~heltai/ Scuola Internazionale Superiore di Studi Avanzati Phone: +39 040 3787 449, Office: 255 -- There are no answers, only cross references. On 05/apr/2010, at 17.52, Andrew McBride wrote: > thanks Michael, that indeed is the other good reason. It's actually the > primary reason. > > Andrew > > Am 05 Apr 2010 um 5:42 PM schrieb Michael Rapson: > >> Hi all, >> >> In my rather brief experience of getting deal.II set up on my laptop >> under OS X (I now dual boot with Ubuntu and run deal.II there mainly >> because it uses less RAM) one of the frustrations that arose was that >> Fortran compilers are not provided. This means that if mpi support for >> PETSc or Trilinos is required then different compilers must be used. >> That leaves the option of compiling gcc yourself or looking for a >> compiled version available on the internet. I ended up compiling >> myself to get a consistent set of compilers to use. >> >> So Fortran support for PETSc and Trilinos is one reason that the xcode >> gcc may not be suitable. If I have followed this post correctly it >> seems like casting problems might be another? >> >> Regards, >> Michael >> >> On Mon, Apr 5, 2010 at 4:50 PM, Andrew McBride <[email protected]> >> wrote: >>> Hi Wolfgang >>> >>> mmm, good question. why spend all this time messing with compilers. Well >>> originally, it was to fix the issue with casting that we dealt with in a >>> previous series of mails. Thus, I switched to the HPC version of gcc, as >>> suggested by Luca. I recall this being the only issue.Perhaps I've >>> forgotten some other very good reason... >>> >>> That said, I've just reinstalled deal.ii using the standard apple xcode gcc >>> (gcc version 4.2.1 (Apple Inc. build 5646) (dot 1)) and all works fine. So >>> step-21, step-5,step-6 and various others are all good. So, now it's not >>> obvious what advantage there is to using a non-standard gcc on the mac. >>> >>> But now I'm a little confused. Khalid was having problem with step-6. If I >>> run step-6 with the standard mac gcc all is fine: >>> .... >>> Cycle 7: >>> Number of active cells: 3860 >>> Number of degrees of freedom: 18353 >>> >>> I presume Khalid is using osx 10.6 and the latest version of xcode. >>> >>> Regards >>> Andrew >>> >>> Am 05 Apr 2010 um 2:43 PM schrieb Wolfgang Bangerth: >>> >>>> >>>>> Wolfgang, you were correct. The problem lies in the latest gcc 4.5 >>>>> version (4.5.0 20100107 (experimental) (GCC)) released by >>>>> http://sourceforge.net/projects/hpc/. I took the code snippet you sent >>>>> me and ran it on the numerous gcc compilers one seems to accumulate when >>>>> one uses a mac. The results are shown below. To summarise: The following >>>>> gcc versions work fine for compiling the code snippet: gcc version 4.2.1 >>>>> (Apple Inc. build 5646) (dot 1) >>>>> gcc version 4.3.3 (GCC) >>>>> gcc version 4.5.0 20090910 (experimental) (GCC). >>>>> But the latest 4.5 version from hpc fails, i.e. >>>>> gcc version 4.5.0 20100107 (experimental) (GCC) >>>>> fails. >>>> >>>> I've opened GCC bug 43648 for this: >>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43648 >>>> >>>> >>>>> This mac related compiler issue is quite frustrating. >>>> >>>> I get the impression that mac os x is tied rather closely to a particular >>>> version of gcc and that using a different version of gcc leads to all >>>> sorts of problems. Does it work to just use the relevant version of xcode, >>>> or what is the reason y'all are playing with different versions of gcc? >>>> >>>> Best >>>> W. >>>> >>>> ------------------------------------------------------------------------- >>>> Wolfgang Bangerth email: [email protected] >>>> www: http://www.math.tamu.edu/~bangerth/ >>>> >>> >>> _______________________________________________ >>> dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii >>> >> _______________________________________________ >> dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii > > _______________________________________________ > dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
