On 10/19/2018 05:32 PM, Stephen Frost wrote:
As you pointed out previously, the current code *doesn't* work, before or after this change, and we clearly need to rework this to move things into libpgcommon and also fix pg_basebackup. Reverting this would at least get us back to having similar code between this and pg_basebackup, and then it'll be cleaner and clearer to have one patch which moves that similar logic into libpgcommon and fixes the missing exceptions for the EXEC_BACKEND case. Keeping the patch doesn't do anything for the pg_basebackup case, and confuses the issue by having these filename-pattern-whitelists which weren't there before and that should be removed, imv.
I don't think just reverting it is really acceptable. The patch was a response to buildfarm breakage, and moreover was published and discussed before it was applied. If you don't like it I think you need to publish a better solution that will not involve reintroducing the buildfarm error. I don't have a strong opinion about the mechanism. The current conversation does seem to me to be generating more heat than light, TBH. But I do have a strong opinion about not having to enable/disable the TAP test in question constantly.
cheers andrew -- Andrew Dunstan https://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
