Build break caused by "

Hi all,

If I’m reading this correctly, it looks like Change #69502 caused some unit test failures in Linux.

http://lab.llvm.org:8011/builders/lldb-x86_64-freebsd/builds/2513
http://lab.llvm.org:8011/builders/lldb-x86_64-freebsd/builds/2512

Thanks,

Vince

Linux

er, FreeBSD

The failing tests are completely different tests. I changed the TestSetValues.py file and the corresponding main.c file and the test it self contained and it runs fine. At most the "TestSetValues.py" could be failing from my change.

Something else seems to be going on.

The expression parser is failing for printf:

(lldb) expr (int)printf("hello\n")
error: call to a function 'printf(char const*, ...)' ('_Z6printfPKcz') that is not present in the target
error: warning: function 'printf' has internal linkage but is not defined
note: used here
error: The expression could not be prepared to run in the target

This seems to be the bug we need to track down. We will take care of that here at Apple.

We think this is due to a clang change. We will be tracking this down shortly.

Greg,

LLVM r221471 broke this. Somebody made bad assumptions about what the dynamic type of a piece of metadata might be.

Sean

After some local reverting, it turns out that no, it didn’t. Curses!

Sean

I have no idea if this is related, but I made r221378 in LLDB yesterday to fix a build breakage as a result of an upstream LLVM change. I’m not sure what LLVM change introduced the break. At first your r221471 sounded like it might be relevant, but that comes after my change, so it couldn’t have been that.

Either way, maybe a red herring, just rang a bell when you mentioned dynamic type of metadata.

The clang commit r221404 broke this. We are working on a fix with the author.

The clang fix has been reverted for now, so the buildbots should all be happy again. Let me know if they aren't.