**** BEGIN LOGGING AT Sun Aug 04 02:59:57 2019
**** BEGIN LOGGING AT Sun Aug 04 13:17:41 2019
**** BEGIN LOGGING AT Sun Aug 04 16:25:25 2019
Aug 04 17:53:33 <seebs> so i have just had a wacky thought. one of the things i noticed while debugging the O_NOFOLLOW stuff is that part of the reason it can be hard to catch things in pseudo is that it keeps automatically fixing stuff. which is *usually* what we want, I think.
Aug 04 17:54:15 <seebs> but i think it might be useful to have a compile flag which makes it stop doing the fastop optimization, and instead, whenever an op would cause a database fixup to happen, return a diagnostic to the client *which causes it to abort with a diagnostic message*.
Aug 04 17:54:46 <seebs> so the first op that comes in with the wrong file type, or a mismatched inode, or something, would instantly abort and say "this path looks wrong, what we have in the database is..."
Aug 04 17:55:06 <seebs> and my guess is that a couple of builds with that would shake out a handful of bugs and probably fix a lot of strange but sporadic behavior.
Aug 04 18:42:12 <khem> seebs: probably a runtime option
Aug 04 18:42:37 <khem> and maybe send them as warnings all the time
Aug 04 18:45:18 <seebs> possible. it's harder in some cases because we also want to send the actual responses. the issue is partially that there's a bit of protocol change; right now, for a lot of ops that are only telling the server thing, the response happens instantly, before the server looks at the database.
Aug 04 22:01:34 <RP> seebs: I'd happily try some builds with such a config
Aug 04 23:10:54 <khem> seebs: if it can collect the responses and inform it might still be useful for post checks
Aug 04 23:11:21 <seebs> it's on the list of things i will get to in my Copious Free Time
Aug 04 23:36:14 <khem> do you need that 25th hr in the day ?
Aug 04 23:36:27 <khem> secrets !!
**** ENDING LOGGING AT Mon Aug 05 03:02:02 2019