https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVypwELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCJxY2hyb21pdW0ud2luL1dpbjcgKDMyKSBUZXN0cy8yNDQxOC9icm93c2VyX3Rlc3RzL1VISmxjbVZ1WkdWeVFuSnZkM05sY2xSbGMzUXVVSEpsY21WdVpHVnlVR3gxWjJsdVVHOTNaWEpUWVhabGNnPT0MCxITTWFzdGVyRmxha2VBbmFseXNpcxgBDA https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVy5AELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCKtAWNocm9taXVtLndpbi9XaW4xMCBUZXN0cyB4NjQvMTU4NTgvYnJvd3Nlcl9zaWRlX25hdmlnYXRpb25fYnJvd3Nlcl90ZXN0cyBvbiBXaW5kb3dzLTEwLTEwNTg2L1JteGhjMmhRWlhKdGFYTnphVzl1UW5KdmQzTmxjbFJsYzNRdVZISnBaMmRsY2xCeWIyMXdkRlpwWVZCc2RXZHBibEJzWVdObGFHOXNaR1Z5DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVyogELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCJsY2hyb21pdW0ud2luL1dpbjcgVGVzdHMgKDEpLzcxMzM4L2Jyb3dzZXJfdGVzdHMvVUd4MVoybHVVRzkzWlhKVFlYWmxja0p5YjNkelpYSlVaWE4wTGtKc2IyTnJWR2x1ZVZCc2RXZHBibk09DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVyqwELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCJ1Y2hyb21pdW0ud2luL1dpbjcgKDMyKSBUZXN0cy8yNDQyMC9icm93c2VyX3Rlc3RzL1JteGhjMmhRWlhKdGFYTnphVzl1UW5KdmQzTmxjbFJsYzNRdVEyOXRiVzl1VTNWalkyVmxaSE5KWmtGc2JHOTNaV1E9DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVywAELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCKJAWNocm9taXVtLndpbi9XaW4xMCBUZXN0cyB4NjQgKEdDRSkvMTc5L2Jyb3dzZXJfdGVzdHMgb24gV2luZG93cy0xMC0xNDM5My9VR3gxWjJsdVVHOTNaWEpUWVhabGNrSnliM2R6WlhKVVpYTjBMa1Y0Y0dGdVpHbHVaMU50WVd4c1VHeDFaMmx1DAsSE01hc3RlckZsYWtlQW5hbHlzaXMYAQw https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVypgELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCJwY2hyb21pdW0ud2luL1dpbjcgVGVzdHMgKDEpLzcxMzQwL2Jyb3dzZXJfdGVzdHMvUm14aGMyaFFaWEp0YVhOemFXOXVRbkp2ZDNObGNsUmxjM1F1UTI5dGJXOXVSbUZwYkhOSlprSnNiMk5yWldRPQwLEhNNYXN0ZXJGbGFrZUFuYWx5c2lzGAEM In the above cases, it seems pretty obvious there's a culprit somewhere in the suggested revision range, yet no confidence score. Regardless of confidence score, if the data point before the suspected data point has a 100% pass rate and the suspected build is flaky, always run try jobs
Similarly, if the previous data point before the suspect is not 100%, do not run try jobs to avoid false positives
https://findit-for-me.appspot.com/waterfall/flake?key=ag9zfmZpbmRpdC1mb3ItbWVyuAELEhdNYXN0ZXJGbGFrZUFuYWx5c2lzUm9vdCKBAWNocm9taXVtLm1hYy9NYWMxMC45IFRlc3RzIChkYmcpLzQ2MTk1L2hlYWRsZXNzX2Jyb3dzZXJ0ZXN0cy9WR0Z5WjJWMFJHOXRZV2x1UkdsemNHOXpaVU52Ym5SbGVIUkdZV2xzYzBsbVNXNVZjMlV1VW5WdVFYTjVibU5VWlhOMAwLEhNNYXN0ZXJGbGFrZUFuYWx5c2lzGAEM In this case, visually it is quite clear there is a culprit. Idea: If a minimum number of preceding data points before the suspected build is 100% stable, and no data point after the suspected build is even slightly stable (e.g. 98%), run try jobs.
Comment 1 by lijeffrey@chromium.org
, Sep 21 2017