On 2018-06-19 17:18:32 -0400, Andrew Dunstan wrote: > On 06/19/2018 04:54 PM, Andres Freund wrote: > > Could we please also change pg_regress' diff invocations? I find it > > really painful to see differences in buildfarm output due to the > > -C3. Locally I've long exported PG_REGRESS_DIFF_OPTS, but that doesn't > > help on the BF.
> It should do if you put it on the build_env stanza of your config file. I > just looked at your animal skink and didn't see any trace of it. What > PG_REGRESS_DIFF_OPTS do you use? > > But of course this won;t help you with other peoples' animals :-) Oh, I was talking about the wider buildfarm, rather than just mine. > One idea I have been playing with is breaking up the reports so that instead > of a large text blob we have a series of files. Allowing different formats > on the web interface for diff files wouldn't be terribly difficult once we > had that - we'd just need to invoke filterdiff or some such. Of course, we > couldn't supply more context that was there in the first place ;-) I mean that would be good - I've way too often spent considerable time looking for errors - but I don't see why that would be related to this change. Given that we've "officially" stopped relying on context diffs, I don't see why we'd not also retire them in pg_regress. Greetings, Andres Freund