Paul, Thanks for the suggestions. Comments interleaved with your reply.
On 01/24/11 11:20, Paul Davis wrote:
Later versions include an entry in the db info blob whether compaction is running. Though you may still get into a bit of a race condition if compaction finishes faster than you can make the second request.
That wouldn't be good. A test failure could be because the program failed, or because the race condition occured. Not too reliable.
Though the return value of the compaction request should also tell you if it's started successfully.
Testing the record code is good (and ought to be done), but it doesn't provide direct evidence that compaction was requested, and allows some failures to go undetected by the test (such as the program not making the request at all).
By the way, the couchdb database that the program-under-test is being run against is created just for the duration of the test and destroyed thereafter, so I can do anything at all to it that might help. I can give it a special or abnormal configuration, for example. In the config, under http_global_handers, I see there's a _replicate entry that is set to {couch_httpd_misc_handlers, handle_replicate_req}. Is it possible for me to set that to my own handler that might just, for example, create or update some document that my test can query to find out if replication was requested?
Thanks, Wayne Conrad
