[Rd] Update on report PR#1145 (with a cross-reference to PR#1026).

gml4410@ggr.co.uk gml4410@ggr.co.uk
Thu, 1 Nov 2001 18:30:41 +0100 (MET)


   The original report was that compiling with gcc2.95.2 gcc and the
related g77 on Irix 6.,5.13 produced an R which wailed its tests with:

>> On entry to DGEMM  parameter number 1 had an illegal value

   To some extend this confused me, as the source only seems to mention
DGEMM in lowercase.  Eventually, after the debug code I added did
nothing, I realized that the R which was build was not using the
src/appl/blas.f source file: it was using the SGI libblas.

   My suspicion is that this is the cause.  The man page for this
mentions there is a C interface for dgemm but that it uses an enumerated
variable for the first 2 parameters in place of character strings.  I
thing things got mixed up somewhere along the line (perhaps with the
number of "_" appended to external symbils?).

   So, I tried configuring while adding "--without-blas" to the config  
options. This resulted in a SIGSEGV in the modreg test.  (See bug report
PR#1026, which also mentions this).

   So, a further try, this time using the compilers, cc anf f77 while  
still specifying "--without-blas" (so that the SGI libblas is ignored).

   This ran through the tests without crashing.  There were a few minor
differences in the print-tests output, but all at the "0 vs.  10**-300"
level, ie: floating point inaccuracies (which I presume are Ok and any
good statistician will ignore?).

==========
comparing `print-tests.Rout' to `./print-tests.Rout.save'
...259c259
< [1]  0.000000e+00 2.225074e-308 2.225074e-308 2.227299e-308
2.447581e-308
---
> [1] 2.002566e-308 2.222849e-308 2.225074e-308 2.225074e-308 2.225074e-308 2.2$
==========

   and 16 occurrences of:

==========
^H: syntax error
========== 
at the end.

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