[01:25] <wgrant> Ohh, whoops.
[01:25] <wgrant> I didn't mangle sequences for empty tables.
[01:30] <wgrant> That explains the recipe failure earlier, too.
[01:30] <wgrant> And the SPPH/BPPH one was just a silly test.
[01:30] <wgrant> Running the full suite with empty tables mangled too.
[08:55] <wgrant> Last test run is 7000 tests short.
[08:56] <wgrant> Just an unfortunate test_disconnectionerror_view_integration, I think.
[09:11] <wgrant> I'm pleasantly surprised at the lack of DB connection fallout.
[09:40] <wgrant> cjwatson: TestBranchView.test_query_count_index_with_subscribers is breaking the subunit stream, I think.
[09:40] <wgrant> I imagine an extra count query appeared or something.
[09:45] <cjwatson> Yeah, probably due to the +recipes change
[09:46] <cjwatson> I thought something like that might happen but it didn't seem worth contortions to fix
[09:46] <wgrant> Agreed.
[09:46] <cjwatson> I'll fix
[09:47] <cjwatson> It's annoying that Storm doesn't cache that, since I believe it's on the same ResultSet object.
[09:49] <wgrant> Storm's only cache is PK -> object.
[09:50] <cjwatson> Yes.  Still annoying :-)
[09:50] <cjwatson> Oh, we actually gained a query somewhere.
[09:50] <cjwatson> I mean, one fewer query than before.
[09:51] <wgrant> Oh.
[09:51] <wgrant> That's... unusual.
[09:51] <wgrant> Possibly there was previously a LIMIT 1 in a tal:condition, and a COUNT(*) in the tal:content?
[09:55] <cjwatson> Two evaluations of view/context/recipes/count in TAL, I believe.
 followed by <tal:recipes/> because TAL lacks else
[19:44] <blr> morning