R devel - Jun 2001 - R-devel on alpha Tru64: cxml/dxml problem (PR#977)

Hi,

I just run a ./configure --enable-R-shlib; make; make check; with a recent
(11.6.2001) R-devel snapshot on Tru64 4.0F (with libcxml):

make check stops in base-Ex.R at:
> ###--- >>> `plot.window' <<<----- Set up World
Coordinates for Graphics
Window>
>       ## alias         help(plot.window)
>
> ##___ Examples ___:
>
> ##--- An example for the use of  `asp' :
> library(mva)
> data(eurodist)
> loc <- cmdscale(eurodist)
Error in La.eigen(Tmat, symmetric = TRUE) :
        On entry to DSYEV  parameter number 8 had an illegal value
Execution halted

The problem is, that by default the accelerated blas library (libdxml or
libcxml) is found by ./configure and gets used (DSYEV resides in this
library) 

I stopped using libdxml/libcxml with R long ago because I had similar
problems (using libdxml on Tru64 4.0E) with some of my libraries, but I
thought it was my fault. The interesting thing is, that some of the dxml
functions are useable, some not.

Now I took a closer look and recognized that libdxml/libcxml are linked
with the DEC/Compaq Fortran runtime libraries:

agebhard@alpha4-stat[R-devel]$ odump -Dl /usr/shlib/libcxml.so

                        ***LIBRARY LIST SECTION***
        Name             Time-Stamp        CheckSum   Flags Version
/usr/shlib/libcxml.so:
        libUfor.so   Nov 15 02:43:32 1999 0xdf165667     0
        libfor.so    Nov 15 02:43:32 1999 0xcfae3688     0
        libFutil.so  Nov 15 02:43:32 1999 0xa19d96f9     0
        libm.so      Apr 12 18:27:39 1999 0xf29b0962     0 osf.1
        libots.so    Apr 12 18:16:38 1999 0x0876eb23     0
        libmach.so   Apr 12 18:18:25 1999 0x5a7ec0dc     0 osf.1
        libc.so      Jun 12 20:14:23 2000 0xbad758ae     0 osf.1

I guess because we are using gcc/g77 and not DEC cc/f77 we cannot use
CXML/DXML libraries.

And indeed, make check succeeds with "--without-blas" added to the
configure options.

Solution:

I would suggest to modify the configure script not to use dxml/cxml on
alpha-tru64 by default.


This was on an alpha EV6 running 

Digital UNIX V4.0F  (Rev. 1229); Fri Jan  5 16:15:45 MET 2001
gcc version 2.95.2 19991024 (release)
g77 version 2.95.2 19991024 (release) (from FSF-g77 version 0.5.25 19991024
(release))

R (the crashing version)  had been linked to the following libraries:

                        ***LIBRARY LIST SECTION***
        Name             Time-Stamp        CheckSum   Flags Version
bin/R.bin:
        libcxml.so   Dec  2 18:32:43 1999 0x472f7fe7     0
        libm.so      Apr 12 18:27:39 1999 0xf29b0962     0 osf.1
        libreadline.so.3 May  5 19:56:37 2000 0x3a7f2031     0
        libz.so.1.1  Feb 23 18:01:02 1999 0x91c64bf3     0
        libncurses.so.4 Feb 23 17:55:19 1999 0xf6374558     0 4.2
        libc.so      Jun 12 20:14:23 2000 0xbad758ae     0 osf.1


Albrecht

--please do not edit the information below--

Version:
 platform = alphaev67-dec-osf4.0f
 arch = alphaev67
 os = osf4.0f
 system = alphaev67, osf4.0f
 status = Under development (unstable)
 major = 1
 minor = 3.0
 year = 2001
 month = 06
 day = 09
 language = R

Search Path:
 .GlobalEnv, package:ctest, Autoloads, package:base

// Albrecht Gebhardt          Tel.: (++43 463) 2700/3118
// Institut fuer Mathematik   Fax : (++43 463) 2700/3198
// Universitaet Klagenfurt    mailto:albrecht.gebhardt@uni-klu.ac.at
// Universitaetsstr. 65       http://www-stat.uni-klu.ac.at/~agebhard
// A-9020 Klagenfurt, Austria 




-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

>>>>> albrecht gebhardt writes:
> Hi,
> I just run a ./configure --enable-R-shlib; make; make check; with a
recent > (11.6.2001) R-devel snapshot on Tru64 4.0F (with libcxml):
> ...
> I guess because we are using gcc/g77 and not DEC cc/f77 we cannot use
> CXML/DXML libraries.
> And indeed, make check succeeds with "--without-blas" added to
the
> configure options.
> Solution:
> I would suggest to modify the configure script not to use dxml/cxml on
> alpha-tru64 by default.
R developers: anyone against that?

If not: Albrecht, can you suggest code which fixes the problem?  Should
we turn off checking for dxml/cxml in general?  I'd rather not special
case a specific platform (or even in combo with a specific compiler).

-k

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._