[PATCH] Fixing a struct layout bug with bitfields before virtual base

[Katya Romanova]

  Hi David,

  I think Paul's assumptions are correct. Two distinct problems can manifest itself via the same assertion message.

  What we tried to fix here is the regression in r169489.
  r169489 was a very big patch, where the bitfield access and bitfield memory layout was re-written. The test that we provided works with the r169488 (and eariler) and fails with the r169489 (and later). I just checked that the testcase that David provided was failing all along.
  Basically, what we are trying to fix here is a new regression introduced in r169489.

  I think it worth filing a different bug using David's testcase.

http://llvm-reviews.chandlerc.com/D2560