Re: BUG #14999: pg_rewind corrupts control file global/pg_control

On Thu, Apr 05, 2018 at 07:04:40AM +0200, TipTop Labs wrote:
> One final word about my original patch, since the commit message for
> 0001 refers to its limitations: I acknowledge that it cannot cover
> situations where the order of processing files is such that the second
> loop detects non-writable files only after it had already done work on
> some others. It was meant to fix the specific problem I ran into
> without breaking the regression tests that ship with the source code
> ("make check"). The main reason that it is not more elaborate is that
> I did not want to submit anything affecting more than a few lines of
> code without prior input from the community.

That definitely makes sense to me.  Reviews and discussions can take a
long time to settle so this method looks right to me.  It is better not
spend too much time on a patch if you unsure of its shape, and much
better to ask for input first before diving more into the details.  This
stuff is really complicated and the devil is in the details.

> So from my perspective I welcome both 0001 and 0002.

OK, thanks for confirming!
--
Michael