[R] building R 2.4.0 on aix

[Prof Brian Ripley]

I've not seen any response to either of these messages, and so am replying 
in an attempt to provide some explanation.  No core R developer has access 
to AIX, nor have we had for some time.  (Most of the recent workarounds 
for AIX have come from the kind efforts of Ei-ji Nakama.)  It does seem 
AIX rivals Darwin (the underlying OS of MacOS X) for being the least alike 
of the Unix-alikes (but fortunately we have Simon Urbanek to explain 
Darwin to us).

We don't know if R 2.4.0 has been built successfully on AIX: we do know 
that no one reported any problems in the alpha/beta/RC test period.  If no 
one tested it, this indicates a lack of support for the R developers from 
the AIX community.  I believe I have seen a report which indicated that R 
2.3.1 was built successfully.

Some more specific comments to Eric Harley:

> configure:26491: checking how to get verbose linking output from f95
> configure:26502: f95 -c  conftest.f >&5
> f95: 1501-218 file conftest.f contains an incorrect file suffix

It seems your 'f95' is not a conventional Fortran compiler.  But you have 
not told us what compiler it is: only if we know the full story can we 
advise against it.

> but make exits early with errors.

Without telling us what they are, the experts here cannot help.


On Wed, 29 Nov 2006, Eric Harley wrote:

> Roy and others,
>
> I'm also trying to build R 2.4.0 on aix 5 (5.3 in my case, on an
> Intellistation Power 285) and I have not had luck yet.  I notice that
> the only mentions of AIX on this list in the past two months are
> people not having luck building it.  So far there don't seem to be any
> responses, and I don't know if it's from lack of information in the
> questions, or because this isn't the right place to ask the questions.

There have been replies, but no responses to those replies.  We often do 
not find out if problems do get resolved.

> Does anyone know a good place to ask such questions?  If this is the
> right place, please let me know what information you need.
>
> I am unfortunately very new to this business of compiling and linking
> on Unix, so I have trouble knowing what information to provide.
>
> Here's my stab at troubleshooting, which is basically trial-and-error:
>
> ===
> tried just to type ./configure (I can always dream, right?  :-) )
> Failed with the following problem in config.log
>
> configure:26491: checking how to get verbose linking output from f95
> configure:26502: f95 -c  conftest.f >&5
> f95: 1501-218 file conftest.f contains an incorrect file suffix
> configure:26508: $? = 1
> configure: failed program was:
> |       program main
> |
> |       end
> configure:26587: WARNING: compilation failed
> configure:26593: result:
> config.log (93%)
>
> ===
> A little playing around shows me that f95 doesn't like to compile
> files ending in .f, so I try doing
> ./configure F77=f77
>
> In this case, the configure script completes successfully.
> Now I try
> make
>
> but make exits early with errors.
> make check
> gives
>
> make: 1254-002 Cannot find a rule to create target
> ../../src/library/base/all.R from dependencies.
> Stop.
>
> and a few other 1254-004 errors.
>
> ===
> There are some examples in
> http://cran.r-project.org/doc/manuals/R-admin.html , particularly in
> the AIX platform notes (C.9 AIX).  However, the second paragraph says
> "The rest of this section is historical, as the default *_LDFLAGS were
> changed in R 2.4.0." so evidently the examples don't apply to the
> current version of R.  I tried some frankenstein combinations of the
> options given in the examples, using the xl*_* compilers since I don't
> have gfortran on my system, but this kind of thing is time-consuming
> when I'm essentially making random changes.
>
> I'm hoping that someone with more experience than I can at least help
> me sort through the "Cannot find a rule to create target" error, and
> that maybe I can proceed somehow one error at a time and finally end
> up with something that works.
>
> BTW, I did try to google the "Cannot find a rule to create target"
> "from dependencies", but I don't get much that I can understand.  One
> message said something about this means make can't find the source
> code.  If this is the problem, how can I figure out what source code
> make is looking for, where it's looking for it, and how to tell it the
> right place to look?  Or perhaps there's something else I'm missing?
>
> Thanks,
> Eric Harley
>
>
> On 11/20/06, Roy Heimbach <royh at hpc.unm.edu> wrote:
>> I'm trying to build R 2.4.0 on aix 5 and not having much luck.  If
>> anyone on the list has built 2.3 or 2.4 on AIX and would be willing
>> pass along some hints, I would be grateful.  So would the research
>> group that's waiting to use R.
>>
>> Thanks,
>> Roy Heimbach
>> --
>> Roy Heimbach <royh at hpc.unm.edu> / 505-277-8348
>> User Services / Center for High Performance Computing
>> University of New Mexico
>>
>> ______________________________________________
>> R-help at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595