Skip to content

changes in configure (Re: Compilation fails (PR#209))

2 messages · Thomas J Vogels, Martin Maechler

Original
Thomas J Vogels
# 
--17.Jun.1999mysteryboxofun
Content-Type: text/plain


Kurt,

  uh, my original suggestion to check the installed xmkmf was directed
to Frank who wasn't very lucky with his Fortran compiler.
Now for AIX, for which I'm actually compiling R at the moment.

"Kurt" == Kurt Hornik <Kurt.Hornik@ci.tuwien.ac.at> writes:
Kurt> AIX is not a supported platform yet.  If you tell me which flags (cpic,
Kurt> fpic, shlib, maybe something extra for producing the main binary) to use
Kurt> for this platform and the NATIVE compilers, I will gladly tell configure
Kurt> how to get things right on AIX.

I'd like to take you up on your offer, especially since I can't create
the configure file out of configure.in.  
Attached is (hopefully) a patchfile that updates configure fo AIX.  It
also patches the Makefile.in in src/library/modreg/src and it contains
ldAIX4 which is a script for the aux (in R-0.64.1) directory.
(I've put the patch at: www.ece.cmu.edu/~tov/public/R-on-aix42.patch)

Something like the following in a clean R-0.64.1 directory should
work:
$ cd R-0.64.1
$ patch -b -u -p0 < ../R-on-aix42.patch
$ chmod +x aux/ldAIX4
$ mkdir build && cd build
$ # force native compilers:
$ CC=cc ../configure --with-f77 --with-x --without-gnome
$ make

This still leaves the problem with R SHLIB.  AIX needs a file with
symbols that are exported from the main program (R.binary).  A "make
install" should store this file in RHOME/etc/main.exp or somwhere
close.  Lemme think about how to work with in installed R.

A final "make test" runs smoothly up to ../../tests/print-tests.R.
There are some diffs with the length of printed doubles, like,
< [4,] 1.7976931348623157e+308
---
So the printed number in AIX is shorter?  Any comments?

Thanks,
  -tom
5 days later
Martin Maechler
# 
If you have a few minutes time to help R development / testing,
*and* if have more than just a binary version of R, please read on..
<....>

    ToV> "Kurt" == Kurt Hornik <Kurt.Hornik@ci.tuwien.ac.at> writes:
    Kurt> AIX is not a supported platform yet.

      <....>

    ToV> A final "make test" runs smoothly up to ../../tests/print-tests.R.
    ToV> There are some diffs with the length of printed doubles, like, <
    ToV> [4,] 1.7976931348623157e+308 ---
      >> [4,] 1.797693134862315708145e+308
    ToV> So the printed number in AIX is shorter?  Any comments?


I have been mainly responsible for  print-tests (and other "make test"s w/ R).
Already in the past we have realized that the C libraries which are
implementing printf() *are* quite different in their behavior, unfortunately.
The above difference is from formatC()'s output, and
formatC uses C-formatting (as its name might suggest..).

There are other places which might give C-library-dependent output.
Before I change  tests/print-tests.R (and ...Rout.save) again
(such that "make test" == "make check" runs to completion on AIX),
I'd be very glad if many of you R-users installers could run

    make check

	 [I think this will only work from the top-level of the
	  Source/Installation tree; i.e. is not available for those who 
	  only installed a binary distribution.
	  Redhat/SuSe Speak: you need SRPMS, not only RPMS]

and report --- to me, not all of R-devel ---
in case you DON'T get the final OK of the line
------------------------------------------------------------------------
Comparing `print-tests.Rout' to `./print-tests.Rout.save' ... OK
------------------------------------------------------------------------
	{{In case you don't get OK, please send me the WHOLE FILE
	  tests/Examples/print-tests.Rout.fail }}

Other problems of "make check" should also be reported, 
probably to R-devel maybe to R-bugs (which forwards to R-devel).

On the other hand, if make check runs to completion,
you get something like a benchmark (with all flaws of every benchmark)
of your computer's speed in running R:

  grep '^Time elapsed:' tests/Examples/base-Ex.Rout

              (almost the last line of tests/Examples/base-Ex.Rout)

will give you the CPU (and total..) Time that was needed to execute 
*all* examples in the R base package.

Thank you in advance!

Martin Maechler <maechler@stat.math.ethz.ch>	http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum SOL G1;	Sonneggstr.33
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1086			<><
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._