R help - Sep 2000 - R on HP-UX, again: modreg.sl fails, 1 arith test fails

I want to report the progress of my attempts to get R running on HP-UX
with gcc: 

Firstly, I could work around the PIC problem by leaving out all the lib stuff
(I rebuilt gcc with --enable-share, but this (alone) did not help), e.g. the
line 

gcc -shared -fPIC  -o ctest.sl ansari.o fexact.o kendall.o ks.o prho.o swilk.o
-L/usr/local/lib  -lg2c
-lm -L/lib/pa1.1 -L/usr/lib/pa1.1 -u main -L/usr/local/qt/lib 
-L/opt/gcc/lib/gcc-lib/hppa1.1-hp-hpux10.20/2.95.2 -L/usr/ccs/bin -L/usr/ccs/lib
-L/opt/gcc/lib -lm

was replaced by 

gcc -shared -fPIC  -o ctest.sl ansari.o fexact.o kendall.o ks.o prho.o swilk.o
-u main

With this I could run the make process to completion. Even the tests w/ loading
these shared
libs work, with the exception of modreg.sl. Calling
> library(modreg)
results in

Error in dyn.load(x, as.logical(local), as.logical(now)) : 
        unable to load shared library
"/usr/local/lib/R/library/modreg/libs/modreg.sl":
  can't open /usr/local/lib/R/library/modreg/libs/modreg.sl
Error in library(modreg) : .First.lib failed

Another problem is the failure of the arith test:

abs(1 - x / asin(sin(x)))

results in values greater then Meps for x=0.53 and 0.56.

Has anybody seen this? Because the arith test is a basic test, this seems to be
rather
specific to my machine/libs.

At least, I have R running at the HP system and can use my 21" monitor,
instead having
to go to our smallish Win95 computer. But with failed arithmetic tests I am not
that
confident in the results I will get ...

Thomas.


============================================================================Thomas
Hoffmann, Institut fuer Halbleiter- und Mikrosystemtechnik, TU Dresden
E-mail: hoffmann at ehmgs2.et.tu-dresden.de
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Thomas Hoffmann wrote:
> With this I could run the make process to completion. Even the tests w/
loading these shared
> libs work, with the exception of modreg.sl. Calling
> 
> > library(modreg)
> 
> results in
> 
> Error in dyn.load(x, as.logical(local), as.logical(now)) : 
>         unable to load shared library
"/usr/local/lib/R/library/modreg/libs/modreg.sl":
>   can't open /usr/local/lib/R/library/modreg/libs/modreg.sl
> Error in library(modreg) : .First.lib failed
> 
To track down the shared library problem you could try this: If you
add BIND_VERBOSE to the flags for shl_load then shl_load may print
something useful, like the name of the symbol it can't find.  The call
to shl_load is in src/unix/hpdlfcn.c, but even though this is a .c
file it is used by inclusion in src/unix/dynload.c, so that is the
file that needs to be recompiled.

> Another problem is the failure of the arith test:
> 
> abs(1 - x / asin(sin(x)))
> 
> results in values greater then Meps for x=0.53 and 0.56.
> 
I believe a few of the tests are a bit too stringent.  THere are two I
believe where the check is for Meps but HP-UX gives 2 * Meps (or
something roughly like that).

One other test fails due to the fact that exp(a large number) returns
DBL_MAX on HP-UX (PARISC at least) rather than IEEE Inf (this may be
adjustable by changing the rounding mode, but I've never looked into
that).  The test wants it to be Inf.  If those are your only problems
I think you are OK.  If there are more failures than those then there
may be more issues.

luke

-- 
Luke Tierney
University of Minnesota                      Phone:           612-625-7843
School of Statistics                         Fax:             612-624-8868
313 Ford Hall, 224 Church St. S.E.           email:      luke at stat.umn.edu
Minneapolis, MN 55455 USA                    WWW:  http://www.stat.umn.edu
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Thomas Hoffmann <hoffmann at ehmgs2.et.tu-dresden.de> writes:
> > library(modreg)
> 
> results in
> 
> Error in dyn.load(x, as.logical(local), as.logical(now)) : 
>         unable to load shared library
"/usr/local/lib/R/library/modreg/libs/modreg.sl":
>   can't open /usr/local/lib/R/library/modreg/libs/modreg.sl
> Error in library(modreg) : .First.lib failed
Hmm. And the file exists and has the proper permissions?
> Another problem is the failure of the arith test:
> 
> abs(1 - x / asin(sin(x)))
> 
> results in values greater then Meps for x=0.53 and 0.56.
> 
> Has anybody seen this? Because the arith test is a basic test, this seems
to be rather
> specific to my machine/libs.
> 
> At least, I have R running at the HP system and can use my 21"
monitor, instead having
> to go to our smallish Win95 computer. But with failed arithmetic tests I am
not that
> confident in the results I will get ...
Um, how much greater are they? On ix86, I get
> x<-(1:10000)/10000
> min(1 - x / asin(sin(x)))
[1] -2.220446e-16
> max(1 - x / asin(sin(x)))
[1] 2.220446e-16


You could try the same test in raw C, and depending on the result
report an error in R, or lose confidence in *anything* computed on
that machine....

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

> > Error in dyn.load(x, as.logical(local), as.logical(now)) : 
> >         unable to load shared library
"/usr/local/lib/R/library/modreg/libs/modreg.sl":
> >   can't open /usr/local/lib/R/library/modreg/libs/modreg.sl
> > Error in library(modreg) : .First.lib failed
> 
> Hmm. And the file exists and has the proper permissions?
$ pwd
/userdisk/usr_local/lib/R/library/modreg/libs
$ ll
total 640
-rwxrwxr-x   1 hoffmann   users       312744 Sep 15 14:06 modreg.sl
> Um, how much greater are they? 
For my:
platform hppa1.1-hp-hpux10.20
arch     hppa1.1             
os       hpux10.20           
system   hppa1.1, hpux10.20  
status                       
major    1                   
minor    1.1   

I get:
             
> x<-(1:10000)/10000
> min(1 - x / asin(sin(x)))
[1] -4.440892e-16
> max(1 - x / asin(sin(x)))
[1] 3.330669e-16

This is approx. 2*Meps.

Thomas.
============================================================================Thomas
Hoffmann, Institut fuer Halbleiter- und Mikrosystemtechnik, TU Dresden
E-mail: hoffmann at ehmgs2.et.tu-dresden.de

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._