Okay, apparently one copy is in content_browsertests, and one copy is in libcontent.so.
Poking through some other similar feature, it looks like may be caused by incorrectly handling the visibility of features::kReporting (and presumably also features::kNetworkErrorLogging) -- other features, like kUkmFeature, are properly defined in only one object file (and referenced by others), but kReporting is defined in several.
I'm going to try adding a NETWORK_EXPORT sort of thing to //services/network to see if it fixes it.
Status update, sure: it works everywhere but fails to build in the windows component build.
ETA, kind of: it depends when I manage to sort that out.
If you'd like it fixed sooner, I can land this with the relevant code disabled on Windows for now.
Thx! I would recommend against a partial fix, since the most vocal failures in the waterwall are Window tests.
Depending on how complicated this gets, and considering the revert has been ruled out above, I think we should consider disabling the affected tests temporarily.
juliatuttle@: can you try the way to fix the revert merge conflict & land the revert the CL instead? At this points, there are lots of tests affected & I think the best course of action is to revert your CL.
Comment 1 by bugdroid1@chromium.org
, Jan 19 2018