Chris Lattner wrote:> Specifically which test is this? It is possible we can just change the > test to not depend on this.Assembler/2004-02-01-NegativeZero.ll ; RUN: llvm-as < %s | llvm-dis | grep -- -0.0 global double 0x8000000000000000 global float -0.0 The .bc made by both Linux and OpenBSD is identical according to an md5 hash.
On Tue, 1 Jul 2008, Edd Barrett wrote:> Chris Lattner wrote: >> Specifically which test is this? It is possible we can just change the >> test to not depend on this. > > Assembler/2004-02-01-NegativeZero.ll > > ; RUN: llvm-as < %s | llvm-dis | grep -- -0.0 > > global double 0x8000000000000000 > global float -0.0 > > The .bc made by both Linux and OpenBSD is identical according to an md5 > hash.If the output of llvm-dis is "0.0" instead of "-0.0", then that is a serious problem, and we need to fix the asmprinter to work around the different in system libraries. The printing is happening in lib/VMCore/AsmWriter.cpp around line 483. -Chris -- http://nondot.org/sabre/ http://llvm.org/
Chris Lattner wrote:> If the output of llvm-dis is "0.0" instead of "-0.0", then that is a > serious problem, and we need to fix the asmprinter to work around the > different in system libraries.Correct. This is the behavior we see. For now wait to see what the OpenBSD dev's say. Thread here: http://marc.info/?l=openbsd-tech&m=121493983219015&w=2 And yes, my test in the first message is b0rk3d :) See the rest of the thread.