[cfe-dev] Clang tests with new pass manager

[David Greene via cfe-dev]

Petr Hosek via cfe-dev <cfe-dev at lists.llvm.org> writes:

> 2. Some Clang tests match the bitcode that's being emitted, but the
> emitted sometimes differs between the legacy and new pass manager.

This concerns me greatly.  I would expect the output to not be dependent
on the pass manager used.  I understand that clang has two separate code
paths to set up pipelines for the old and new pass managers.  Are these
expected to be functionally equivalent?  If not, why not?

                          -David