On Jun 19, 2013, at 15:58 , Berend Hasselman wrote: > > On 19-06-2013, at 14:17, peter dalgaard <pda...@gmail.com> wrote: > >> >> >> >> Thanks. I think I have it nailed down now. The culprit was indeed in our >> reference BLAS (I had only checked the LAPACK code), cmplxblas.f to be >> specific. Revision 53001 had a number of IF statements being commented out, >> but two of the changes looked like this: >> >> @@ -1561,7 +1561,7 @@ >> C( J, J ) = DBLE( C( J, J ) ) >> END IF >> DO 170 L = 1, K >> - IF( ( A( J, L ).NE.ZERO ) .OR. ( B( J, L ).NE.ZERO ) ) >> +c IF( ( A( J, L ).NE.ZERO ) .OR. ( B( J, L ).NE.ZERO ) ) >> $ THEN >> TEMP1 = ALPHA*DCONJG( B( J, L ) ) >> TEMP2 = DCONJG( ALPHA*A( J, L ) ) >> >> Notice that the continuation line was NOT commented out. So FORTRAN, being >> what it is, continues the line before the comment and parses it as >> DO 170 L = 1, KTHEN >> with KTHEN uninitialized! and things go downhill from there. >> >> (The uninitialized variable was actually hinted at in PR14964 and the fact >> that I could get one of my builds to segfault also helped.) >> > > > And it seems that line 1536 and 1537 ( R-3.0.1 source) have a similar change, > so that will also have be put right. > > Berend > >
Yes, that's the other one of the two changes. -pd -- Peter Dalgaard, Professor Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd....@cbs.dk Priv: pda...@gmail.com ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
