Re: BUG #4879: bgwriter fails to fsync the file in recovery mode

Simon Riggs <simon(at)2ndQuadrant(dot)com> writes:
> On Fri, 2009-06-26 at 16:48 -0400, Tom Lane wrote:
>> * I find the RecoveryInProgress test in XLogNeedsFlush rather dubious.
>> Why is it okay to disable that? For at least one of the two callers
>> (SetHintBits) it seems like the safe answer is "true" not "false".
>> This doesn't matter too much yet but it will for hot standby.

> IIUC you think it is safest to *not* write hint bits during Hot Standby?

Well, the point I was trying to make is that if you want the code to do
nothing then hot-wiring it to return "false" all the time during hot
standby doesn't accomplish that. I think it probably is safe to update
hint bits during HS, but we need to think through the timing relative to
WAL processing and make sure that we don't set them too early.

Maybe that analysis has actually been done already, but there's no
evidence of it in the code or comments.

regards, tom lane