> Cons: You're mucking around in important root directories with root
> privileges (without sudo you can't redirect links), so mistakes can be
> harmful to your system. Therefore, choice 2 requires care and confidence in
> what you're doing.
So.... Don't install it as root. You don't need to be root to
install system-wide. Very few programs actually need root privileges,
and some that say they do probably shouldn't.
Somewhere back in the reflector, I gave details about how to do this,
system-wide, but not as root. You might take a look.
On Wed, Aug 25, 2010 at 4:10 PM, Jason Swails <jason.swails.gmail.com> wrote:
> Hello,
>
> There are a couple avenues open to you:
>
> 1) First, edit the configure script and replace "gcc" with "gcc44" wherever
> you can find it (or if you can confidently pick out the important spots,
> just replace it where you need to). Same with gfortran -> gfortran44 and
> g++ -> g++44.
>
> 2) An alternative option is to create symbolic links to gfortran44, gcc44,
> and g++44 named gfortran, gcc, and g++, respectively, to a place in your
> PATH *BEFORE* the default gcc-4.1.2 binaries exist. I'm not exactly sure
> how RHEL works, but for Ubuntu, /usr/bin/gcc is just a symbolic link to
> whichever version it uses (i.e. /usr/bin/gcc-4.4). If this is the case, you
> can just redirect those links to the version you want to use. Note, a
> symbolic link is probably safest here (to preserve any relative paths that
> may be built into the binary). Do this with "ln -s" as opposed to just
> "ln". For example:
>
> ln -s /path/to/gcc44 /new/path/to/gcc
>
> where /new/path/to comes before /usr/bin in $PATH.
>
> For my systems, I use (maybe a variant) of option (2). But here are the
> pros and cons of each, IMO:
>
> 1) Pros: It works and it's an easy change that can be implemented with a
> single find-and-replace command with your favorite text editor.
> Cons: It's a fix for amber only. If another package you want to build needs
> to be compiled with the newer compilers, you'll have to figure out how to
> make the analogous fix to that package as well. Also, if you try to use
> that configure script on another machine (i.e. if you copy over the source
> code without untarring a fresh copy if you've made modifications or
> something), you'll need to remember to undo those changes or it won't work
> (unless of course it's an equivalent RHEL machine).
>
> 2) Pros: Once the change is made, your default compilers for everything will
> be the 4.4 versions.
> Cons: You're mucking around in important root directories with root
> privileges (without sudo you can't redirect links), so mistakes can be
> harmful to your system. Therefore, choice 2 requires care and confidence in
> what you're doing.
>
> As for using the correct libraries: each version of the gcc compiler
> *should* link the proper libraries without having to do anything yourself,
> and that has been my experience. But then again, I've always been careful
> to use soft links to compilers when taking route 2 so if the compilers rely
> on a built-in relative path to its libraries, referencing the link won't
> break that.
>
> It's been verbose, but hopefully helpful.
>
> All the best,
> Jason
>
> On Wed, Aug 25, 2010 at 3:51 PM, Richard Owczarzy <rowczarzy.idtdna.com>wrote:
>
>> Hello Mark,
>>
>> I am strugling with the same bug. Could you be more specific about solution
>> for RHEL5? I can install gcc version 4.4.0 using yum packager alongside
>> standard 4.1.2 version. However, new version have executables gcc44, g++44
>> and gfortran44. How do I modify Amber 11 configure or config.h files to
>> properly use new gcc version? How do I point to correct libraries for gcc
>> version 4.4?
>>
>> Thank you,
>>
>> Richard Owczrazy
>> Integrated DNA Technologies
>>
>>
>>
>> -----------------------------------------------------------------------------------------------------------
>> From: Mark Williamson <mjw.sdsc.edu <mailto:mjw.sdsc.edu?Subject=Re%3A%20%5BAMBER%5D%20Workaround%20for%20gfortran-4.1.2%3F>
>> >
>> Date: Tue, 20 Jul 2010 16:51:43 -0700
>>
>>
>> Marc Moreau wrote:
>> > Hello all,
>> >
>> > I am trying to compile against gcc-4.1.2 which includes the buggy
>> > gfortran-4.1.2.
>> >
>> > I'm trying to compile with -O2 instead of the default -O3. Does anyone
>> > know of any other things to try, or specific compiler flags that allow
>> > a work-around of the buggy section in fortran?
>> >
>> > I am aware that 4.3.2 and 4.4.x are out, but my system is old ( Centos
>> > 4.3 ) which doesn't have RPMS for those compilers. Further, if I
>> > upgrade the compiler ( more specifically the libraries ) I need to go
>> > back and recompile && regression test all the other MD suites in use.
>> > It's a non-trivial number so I really don't want to do that. We
>> > expect that we will be getting new Hardware soon which will allow for
>> > an upgrade of OS and such, then I can roll the old hardware to the new
>> > system. But that is later. For now, upgrade is not an option.
>> >
>> >
>> > -- Marc
>> >
>>
>> Hi Marc,
>>
>> I've hit this frustrating issue in the past with RHEL4 (which
>> essentially is what CentOS4 cloned). There is a RPM based solution
>> within RHEL5 and CentOS5, but I also had upgrade constraints here.
>>
>> My solution for RHEL4 was to compile the 4.4 gcc compiler from scratch
>> and then use it to build AMBER. I've written a howto on this and you can
>> find it at:
>>
>> http://www.wmd-lab.org/mjw/howtos/compiling_gcc_4_4_4_under_RHEL4.txt
>>
>> It works rather well for me.
>>
>>
>> --
>> Mark Williamson, Post Doc
>> Walker Molecular Dynamics Group
>> Room 395E
>> San Diego Supercomputer Center
>> 9500 Gilman Drive
>> La Jolla, CA 92093-0505
>> Email: mjw at sdsc.edu
>> Office: 858-246-0827
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
:-) Lachele
Lachele Foley
CCRC/UGA
Athens, GA USA
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 25 2010 - 15:30:03 PDT
