[llvm-commits] [llvm] r115393 - in /llvm/trunk: CMakeLists.txt lib/Target/MSP430/InstPrinter/CMakeLists.txt lib/Target/MSP430/InstPrinter/MSP430InstPrinter.cpp lib/Target/MSP430/InstPrinter/MSP430InstPrinter.h lib/Target/MSP430/InstPrinter/Makefi

[Óscar Fuentes]

Nick Lewycky <nlewycky at google.com> writes:

> Okay. You can see that almost all of the open-source builders were broken:
>
>   http://google1.osuosl.org:8011/console
>
>  <http://google1.osuosl.org:8011/console>in that time. It's impossible for
> this particular error to occur in a cmake build because cmake doesn't run
> find-cycles.pl (last i checked).

Yes, it does, but it seems there is a bug on how cmake runs it, so when
find-cycles.pl complains it is silenced.

> My suspicion is that the cmake builders were working fine while
> configure+make ones were not?

AFAIK, there is only one cmake buildbot (which runs Windows.) No doubt
that Jim is talking about he internal Apple bots that uses the
configure&make build.

> I'm going to wind back to the broken point and try to reproduce the failure
> and see if I can figure out what the cyclic dependency actually was.