Skip to content

Version 1.5.1 failing make check.

4 messages · Rolf Turner, Peter Dalgaard, Thomas Lumley +1 more

Original
Rolf Turner
# 
I just tried to install R version 1.5.1, and it is failing
``make check'', as follows:

$ make check
`Makedeps' is up to date.
running code in 'base-Ex.R' ...*** Error code 1
make: Fatal error: Command failed for target `base-Ex.Rout'
Current working directory /home1/faculty/rolf/Traal/Rtmp/R-1.5.1/tests/Examples
*** Error code 1
make: Fatal error: Command failed for target `test-Examples'
Current working directory /home1/faculty/rolf/Traal/Rtmp/R-1.5.1/tests/Examples
*** Error code 1
make: Fatal error: Command failed for target `test-Examples'
Current working directory /home1/faculty/rolf/Traal/Rtmp/R-1.5.1/tests
*** Error code 1
make: Fatal error: Command failed for target `test-all-basics'
Current working directory /home1/faculty/rolf/Traal/Rtmp/R-1.5.1/tests
*** Error code 1
make: Fatal error: Command failed for target `check'

Amongst other things:

	(1) in tests/Examples/base-Ex.Rout.fail we find

		Error: !is.nan(var(1)) is not TRUE

	(2) in tests/ a diff of arith-true.Rout.fail and
	arith-true.Rout.save gives

			80c80
			< [1] FALSE
			---
			> [1] TRUE
			82c82
			< [1] FALSE
			---
			> [1] TRUE

	Looking in these files shows the difference to come from

		all(!is.nan(c(1,NA)))
	and
		all(c(F,T,F) == is.nan(list(1,NaN,NA)))

	I checked that in R version 1.4.1, which is currently running
	on our system, these give TRUE, in agreement with
	arith-true.Rout.save.  Likewise

		!is.nan(var(1))

	gives TRUE, as it apparently should, under 1.4.1.

	(So the problem would seem --- ??? --- to be with the new
	version rather than with our system.)

	There are a bunch of other errors as well.

Should I just ignore the fact that ``make check'' fails, and go ahead
and install 1.5.1?  (I ***really*** don't like the idea of doing
this!)  If not what, if anything, can I do to make it not fail?

Version information (when I fire up the new, but not installed,
version 1.5.1, by invoking bin/R:

platform sparc-sun-solaris2.7
arch     sparc               
os       solaris2.7          
system   sparc, solaris2.7   
status                       
major    1                   
minor    5.1                 
year     2002                
month    06                  
day      17                  
language R     

Thanks for any wisdom and insight that anyone can provide.

					cheers,

						Rolf Turner
						rolf at math.unb.ca

P. S.  I ***think*** that version 1.5.1 was made with a different
gcc than was 1.4.1.  Could this be the problem?  And if so, what
do I do about it?

						R. T.
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Peter Dalgaard
# 
Rolf Turner <rolf at math.unb.ca> writes:
....
...
Didn't we just do that one and traced it to a compiler issue? You
might want to check the mailing list archives. It's certainly possible
to get R running properly on Sol.2.7 -- both Brian Ripley and I still
install on such systems.
Thomas Lumley
#
On Tue, 3 Sep 2002, Rolf Turner wrote:
<snip> make: Fatal error: Command failed for target `check'
<snip>
Well, an interaction between them. Certainly make check works on a lot of
systems with R 1.5.1.  It wouldn't get released otherwise.
No, that would be a really bad idea.
<snip>
Looking at the forthcoming 1.6.0 I see that
   o   Workaround for optimization bugs on gcc 3.1/2 on 32-bit Solaris.
is listed in NEWS, so it might be worth either

a) reducing the optimisation level by editing CFLAGS in config.site
or
b) trying the pre1.6.0 code to see if the problem has gone away.


	-thomas

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
1 day later
Brian Ripley
# 
On Solaris 7/8 with gcc 3.1 or later and a 64-bit kernel you need to
compile arithmetic.c without optimization.

It's a compiler bug.  There is a workaround in 1.6.0.  It does not occur
with a 32-bit kernel, nor with SunPro cc, nor with gcc 3.0.4.
On Tue, 3 Sep 2002, Thomas Lumley wrote: