Right now, ReportProcessor::resolve_build_id fetches all builds that match the builder number and time. This can result in a long delay & a high memory usage.
Created attachment 331600 [details] Fixes the bug
Comment on attachment 331600 [details] Fixes the bug rs=me. I'm a bit curious why the maximum limit of 2 builds is hard-coded. Is this hardware-dependent, and "2" is the most reasonable cap for our machines we're using to run the perf dashboard? It seems like maybe it should be configurable, but I'm not familiar with this code at all :P
Comment on attachment 331600 [details] Fixes the bug Clearing flags on attachment: 331600 Committed r227166: <https://trac.webkit.org/changeset/227166>
All reviewed patches have been landed. Closing bug.
(In reply to Wenson Hsieh from comment #2) > Comment on attachment 331600 [details] > Fixes the bug > > rs=me. > > I'm a bit curious why the maximum limit of 2 builds is hard-coded. Is this > hardware-dependent, and "2" is the most reasonable cap for our machines > we're using to run the perf dashboard? It seems like maybe it should be > configurable, but I'm not familiar with this code at all :P It's two because we're trying to check that it's unique. When there are more than two entries, we know there is at least one duplicate. We don't need to fetch all the duplicates to know that there are some duplicate.
<rdar://problem/36635350>