Re: pg_upgrade from 9.4 to 10.4

On Tue, Jul 31, 2018 at 06:29:34PM -0400, Tom Lane wrote:
> Bruce Momjian <bruce(at)momjian(dot)us> writes:
> > On Tue, Jul 31, 2018 at 03:26:47PM -0400, Tom Lane wrote:
> >> This patch evidently broke build farm member jacana, although only
> >> in the 9.3 branch. It's been failing with
> >> Jul 28 23:22:30 The system cannot find the path specified.
>
> > Well, that's interesting. I have a test script that upgrades all
> > version combinations of Postgres from 9.3 to head and it worked for me.
> > However, I now see that the tests for checking a live server were wrong,
> > and 9.3 must have a slightly different test than the others. Fixes
> > applied. Thanks for finding this so quickly.
>
> After poking around a bit more, I think the more likely explanation is
> that 9.3 was still using the SYSTEMQUOTE macro in commands to be popen'd,
> and you did not make that adjustment when back-patching into that branch.
> Note the adjacent pre-existing invocation of pg_controldata:
>
> snprintf(cmd, sizeof(cmd), SYSTEMQUOTE "\"%s/%s \"%s\"" SYSTEMQUOTE,
> cluster->bindir,
> live_check ? "pg_controldata\"" : "pg_resetxlog\" -n",

Oh, jacana must be a Windows server with spaces in the file name paths.
Fixed. Thanks again. I really didn't want to backpatch the original
fix but had to since it could produce corrupt upgrades.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ As you are, so once was I. As I am, so you will be. +
+ Ancient Roman grave inscription +