diff options
| author | Tianjie Xu <xunchang@google.com> | 2018-08-06 12:59:23 -0700 |
|---|---|---|
| committer | Tianjie Xu <xunchang@google.com> | 2018-08-16 15:33:09 -0700 |
| commit | 2a0ea6321578d30be3caa806a65149bffd09672a (patch) | |
| tree | 5c9ad105ff6f67d1c501a050bea77002f94aa3be /metrics_utils.h | |
| parent | b67b90aa766fc41eb5d889d8c5fdc6a7089f09b7 (diff) | |
Fix two android metrics reporting in update engine
1. The AttemptCurrentBytesDownloadedMiB was reported in bytes.
2. We were using the same start timestamp (in monotonic_time) for both
boot_time and monotonic_time. This leads to an inaccurate update
duration reporting. This cl saves an extra perf for the boot time and
calculates the duration accordingly.
Also report these information with a different metrics name.
Bug: 30989466
Test: Run an update and check the report
Change-Id: I69e20648501a736a59ff60ca9202ce7bdd9b6e34
Diffstat (limited to 'metrics_utils.h')
| -rw-r--r-- | metrics_utils.h | 8 |
1 files changed, 7 insertions, 1 deletions
diff --git a/metrics_utils.h b/metrics_utils.h index d08cc4a7..8f1aad1e 100644 --- a/metrics_utils.h +++ b/metrics_utils.h @@ -87,10 +87,16 @@ void SetPayloadAttemptNumber(int64_t payload_attempt_number, // Persists the finished time of an update to the |kPrefsSystemUpdatedMarker|. void SetSystemUpdatedMarker(ClockInterface* clock, PrefsInterface* prefs); -// Persists the start time of an update to |kPrefsUpdateTimestampStart|. +// Persists the start monotonic time of an update to +// |kPrefsUpdateTimestampStart|. void SetUpdateTimestampStart(const base::Time& update_start_time, PrefsInterface* prefs); +// Persists the start boot time of an update to +// |kPrefsUpdateBootTimestampStart|. +void SetUpdateBootTimestampStart(const base::Time& update_start_boot_time, + PrefsInterface* prefs); + // Called at program startup if the device booted into a new update. // The |time_to_reboot| parameter contains the (monotonic-clock) duration // from when the update successfully completed (the value in |