Since we have test archives uploaded to the results database, we should add an endpoint to serve the contents of those archives.
Created attachment 378482 [details] Patch
Created attachment 378692 [details] Patch
Created attachment 378994 [details] Patch
Created attachment 385813 [details] Patch
Originally, there were some significant performance problems with this approach, but now that we've de-duplicated archives (in https://bugs.webkit.org/show_bug.cgi?id=204860) this ends up working pretty well. It's not as fast as serving content already on disk, but it's fast enough to be usable, and it will definitely beat out our Internal buildbot instance which is under some pretty substantial load.
Created attachment 385980 [details] Patch
Comment on attachment 385980 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=385980&action=review > Tools/resultsdbpy/resultsdbpy/model/archive_context.py:237 > + archive_list.append(item) So do we not want to append master to the list if it is the master?
Comment on attachment 385980 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=385980&action=review >> Tools/resultsdbpy/resultsdbpy/model/archive_context.py:237 >> + archive_list.append(item) > > So do we not want to append master to the list if it is the master? Correct, we only want to append the "master" to the list once we determine that we're actually unmastered. And actually, I can even see why this worked! line 242 should also be removed.
Created attachment 385998 [details] Patch
Comment on attachment 385998 [details] Patch Clearing flags on attachment: 385998 Committed r253721: <https://trac.webkit.org/changeset/253721>
All reviewed patches have been landed. Closing bug.
<rdar://problem/58057657>