pw_contest: limit the number of updates on the first run

We seem to report 12 checks for the first branch that contains
a patch, because the test case count grows slowly. This happens
mostly for branches which have failures, as failure is declared
as soon as one remote fails, even if others are pending.

So limit the updates of test count for otherwise equivalent state
if some remotes are still running.

Signed-off-by: Jakub Kicinski <[email protected]>

Author: kuba-moo

pw_contest.py

@@ -124,6 +124,7 @@ def branch_summarize(filters: dict, results_by_branch: dict) -> dict:
     for name, branch in results_by_branch.items():
         code = 0
         test_cnt = 0
+        pending = 0
         for remote in filters["remotes"]:
             new_code = Codes.PENDING
             if remote in branch:
@@ -132,7 +133,11 @@ def branch_summarize(filters: dict, results_by_branch: dict) -> dict:
                     new_code = branch[remote]['code']
                 test_cnt += branch[remote]["cnt"]
             code = max(code, new_code)
+            if new_code == Codes.PENDING:
+                pending += 1
         summary[name] = {'result': code_to_str[code], 'code': code, 'cnt': test_cnt}
+        if pending:
+            summary[name]['pending'] = pending
     return summary
 
 
@@ -153,7 +158,11 @@ def result_upgrades(states: dict, item_id: str, outcome: dict, branch: str):
     if prev['code'] > outcome['code']:
         return True
     if prev['code'] == outcome['code']:
-        return prev['cnt'] < outcome['cnt']
+        # use the run with most reported results, but wait for it to finish
+        # otherwise first contest for a patch updates the check every time
+        # a remote finishes and bumps the test case count (~12 updates)
+        if prev['cnt'] < outcome['cnt']:
+            return bool(outcome.get('pending'))
     return False