I'm seeing this when I run the fibonacci program via the win32 build.
VRegInfo has size 0, and the debug STL in VStudio is asserting when
&VRegInfo[0] is hit. Is anyone else seeing this?
unsigned createVirtualRegister(const TargetRegisterClass *RegClass) {
assert(RegClass && "Cannot create register without
RegClass!");
// Add a reg, but keep track of whether the vector reallocated or
not.
void *ArrayBase = &VRegInfo[0];
...
I'm going to try reproing from my mac build, but if any of you are
sync'ed to the trunk and could compile and run fibonacci, I'd appreciate
it.
Thanks,
Chuck.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20080102/5fa3f7da/attachment.html>
On Wednesday 02 January 2008 17:33, Chuck Rose III wrote:> I'm seeing this when I run the fibonacci program via the win32 build. > > VRegInfo has size 0, and the debug STL in VStudio is asserting when > &VRegInfo[0] is hit. Is anyone else seeing this?People will probably see it if they enable GLIBCXX_DEBUG (--enable-expensive-checks to configure). Check the list archives (-dev and -commits) to see how I've handled these kinds of problems. We really need to be doing some testing with this enabled. I've had completely clean checkouts broken again by updating from svn and getting someone's changes that reintroduced bad code. I ALWAYS test in debug mode with this enabled. Building llvm-gcc and running the tests will catch these problems and it doesn't take much longer than a debug build without the expensive checks. -Dave