[llvm-commits] buildbot failure in llvm on llvm-x86_64-linux-vg_leak

[Chris Lattner]

On Apr 4, 2010, at 2:26 PM, Jakob Stoklund Olesen wrote:

>> 
>> // FIXME: This should use ManagedStatic to manage the pass registrar.
>> // Unfortunately, we can't do this, because passes are registered with static
>> // ctors, and having llvm_shutdown clear this map prevents successful
>> // ressurection after llvm_shutdown is run.
>> static PassRegistrar *getPassRegistrar() {
>> static PassRegistrar *PassRegistrarObj = 0;
>> 
>> 
>> The code is intentionally leaking the PassRegistrar object.  I don't know what the right solution is here.
> 
> Bizarre. Why is only one test failing?

More bizzarre: why did the next build just pass?  I suspect that this is some integer that looks like a pointer, causing valgrind to think that a datastructure is alive.

-Chris