debuginfo: Activate LLDB tests on Darwin.

[michaelwoerister]

Now that there are build bots with a stable enough LLDB version on OSX we can finally let the tests run on every PR!
😹

[michaelwoerister]

cc @alexcrichton

[alexcrichton]

Awesome work @michaelwoerister!

[michaelwoerister]

Whoops, it seems that I have broken GDB tests depending on breakpoints at specific line numbers. I'll fix them when I get to a computer...

On October 16, 2014 5:31:41 PM CEST, Alex Crichton notifications@github.com wrote:

Awesome work @michaelwoerister!

---

Reply to this email directly or view it on GitHub:
#18092 (comment)

Sent from my Android device with K-9 Mail. Please excuse my brevity.

[michaelwoerister]

The tests should be fixed now.

[michaelwoerister]

I'm not sure what went wrong the last time. It said that it couldn't create the debugger 'target', meaning that it probably didn't find the executable to debug (or couldn't access it because of access rights or code signing issues). Interestingly it worked on the try branch just now, without any change. Both times it ran on bot mac6.

I've added some more output to the LLDB script, so if it fails again, I should hopefully be able to find out what the actual problem was.

[michaelwoerister]

Alright! So the tests were actually executed two times, once on mac5 and once on mac6. On mac4 they have been ignored as they should be.
I'm a little worried that testing got a bit less deterministic. In theory it may be possible that the tests are not run at all for a PR because testing only runs on machines where LLDB is too old. That could lead to a regression being introduced but showing up only later for PRs that don't have anything to do with the problem. We'll see if that becomes a problem in practice.
Thanks, Alex, for all the help with figuring out what to do with the build system!

[alexcrichton]

I think that we're guaranteed at least for now that at least one build will run on mac5 or mac6 (b/c mac4 can't run all the builds at once), so we should be somewhat ok on that front (hopefully not too bad!).

Awesome work @michaelwoerister, so glad to see these tests checked in!