Roberto, I didn't have time this morning for the full make check in ppl cvs to run. However I ran it through the float and double sections containing those previously failing tests and they now seem to pass fine on i686-apple-darwin9. I did have to compile ppl cvs with Apple's gcc-4.2 compiler. The current ppl cvs code fails to compile with Apple's stock gcc 4.0.1 compiler. I'll post the error message for that later tonight. Is there a schedule for when ppl cvs will be released as the next version of ppl? Also, have the APIs changed enough between 0.9 and cvs that cloog-ppl from the git won't compile against ppl cvs? I am considering trying to use ppl cvs to build cloog-ppl git and graphite to see if those two failing graphite tests I saw disappear with ppl cvs. Jack
On Thu, Aug 07, 2008 at 08:22:58AM +0200, Roberto Bagnara wrote:
Jack Howarth wrote:
I have verified that there are no make check failures in ppl-0.9 on powerpc-apple-darwin9 with Xcode 3.1. The problems appear to be specific to Macintel. I see the same 5 failures when ppl-0.9 is built with Apple's gcc 4.0.1 or gcc 4.2 compilers (so it is unlikely to be a compiler bug). The failing test cases output is appended below.
Jack,
I will check more carefully, but I believe you should not worry about these failures: they are imperfections of the testsuite (which we will fix, if not already fixed in CVS), not of the library. Your attempts to use the CVS HEAD version of the PPL make a lot of sense anyway. Cheers,
Roberto-- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it