Fetch the data from last run earlier

The run command, before bisect is used, tries to get rid of tests
that always fail - to do so it generates new runs, which loses
the reference to the last one before bisect was requested.

To fix this, fetch the last run earlier in the run code.
On bisect side, run all candidate tests, not only failing ones.
And drop an extra comma which broke things.

This version of bisect only find conflicting tests within a
worker since the bisect code relies on the ordering of tests in
the database for a worker.

stestr/bisect_tests.py

@@ -31,7 +31,7 @@ def __init__(self, latest_run, conf, run_func, repo, test_path=None,
         self.repo_type = repo_type
         self.repo_url = repo_url
         self.serial = serial
-        self.concurrency = concurrency,
+        self.concurrency = concurrency
         self.test_path = test_path
         self.top_dir = top_dir
         self.run_func = run_func
@@ -57,7 +57,7 @@ def bisect_tests(self, spurious_failures):
                     repo_type=self.repo_type, repo_url=self.repo_url,
                     serial=self.serial, concurrency=self.concurrency,
                     test_path=self.test_path, top_dir=self.top_dir)
-                self.run_func(cmd, True, True, False, False,
+                self.run_func(cmd, False, True, False, False,
                               pretty_out=False,
                               repo_type=self.repo_type,
                               repo_url=self.repo_url)

stestr/commands/run.py

@@ -318,6 +318,11 @@ def run_tests():
                               stdout=stdout,
                               abbreviate=abbreviate)
     else:
+        # Where do we source data about the cause of conflicts.
+        # XXX: Should instead capture the run id in with the failing test
+        # data so that we can deal with failures split across many partial
+        # runs.
+        latest_run = repo.get_latest_run()
         # Stage one: reduce the list of failing tests (possibly further
         # reduced by testfilters) to eliminate fails-on-own tests.
         spurious_failures = set()
@@ -348,11 +353,6 @@ def run_tests():
         if not spurious_failures:
             # All done.
             return 0
-        # Where do we source data about the cause of conflicts.
-        # XXX: Should instead capture the run id in with the failing test
-        # data so that we can deal with failures split across many partial
-        # runs.
-        latest_run = repo.get_latest_run()
         bisect_runner = bisect_tests.IsolationAnalyzer(
             latest_run, conf, _run_tests, repo, test_path=test_path,
             top_dir=top_dir, group_regex=group_regex, repo_type=repo_type,