Allow non-fatal timestamp solver failures
This refactors the timestamp filters so that it is possible to observe
timestamp solver failures without having the application crash.
This also makes it so that the timestamp_extractor code gets at least
somewhat exercised by our tests.
The only behavioral change this should cause for normal users is that
the stack trace will look slightly different when errors occur, because
the LOG(FATAL) now occurs when evaluating the result of QueueNextTimestamp()
instead of deep in the code (the prior LOG(FATAL)'s are replaced by
LOG(ERROR)'s to keep them visible).
Change-Id: I01e4de8df724c2853cc62ac588668b811563b33d
Signed-off-by: James Kuszmaul <james.kuszmaul@bluerivertech.com>
diff --git a/aos/network/testing_time_converter.h b/aos/network/testing_time_converter.h
index 20b8e16..ae85cee 100644
--- a/aos/network/testing_time_converter.h
+++ b/aos/network/testing_time_converter.h
@@ -41,8 +41,8 @@
void AddNextTimestamp(distributed_clock::time_point time,
std::vector<logger::BootTimestamp> times);
- std::optional<std::tuple<distributed_clock::time_point,
- std::vector<logger::BootTimestamp>>>
+ std::optional<std::optional<std::tuple<distributed_clock::time_point,
+ std::vector<logger::BootTimestamp>>>>
NextTimestamp() override;
void set_boot_uuid(size_t node_index, size_t boot_count, UUID uuid) {