Would a (temporary) solution to spurious CI failures be to just run the test a second time if the first one fails? It would occasionally slow down getting results for good code, and it's a pretty ugly solution, but it would reduce the false positives from maybe once every 50 tries to once every 2500 (assuming no correlation between individual attempts).
Comparison tool on Travis failing: run again? #4923
issue sipa opened this issue on September 15, 2014-
sipa commented at 4:55 PM on September 15, 2014: member
-
-
theuni commented at 5:13 PM on September 15, 2014: member
yep, we discussed this last night on irc and agreed on this exact compromise. travis has a built-in 'travis-retry' that will try a step 3 times before failure. This makes sense as a temporary fix for us. Will PR it.
-
laanwj commented at 9:07 AM on September 16, 2014: member
Sounds good to me.
- gavinandresen closed this on Sep 16, 2014
- MarcoFalke locked this on Sep 8, 2021
Linked (view graph)