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

On Fri, Mar 02, 2018 at 04:35:07PM +0100, Dmitry Dolgov wrote:
> It maybe a stupid question, but why do you need to reset errno here? Is it to
> avoid its value being carried from previous calls of `open_target_file`? In
> this case if you put the code with `errno == EACCESS` under the if condition
> `if (dstfd < 0)`, then as far as I remember you should always get relevant
> errno. At the same time maybe it's valid to reset `errno` before `open`, like
> with `getpriority`:

[ ... reads and feels stupid ...]

Of course all the checks should be where dstno is negative...

I have done a second pass on the patch, and attached is a new version.
This fixes this handling of errno and addresses some typos. I have also
fixed the test case where one of the read-only files was not properly
switched to do so. I have also added a commit log message.
--
Michael