Re: pg_upgrade failure on Windows Server

On Fri, Mar 20, 2015 at 11:14 PM, Asif Naeem <anaeem(dot)it(at)gmail(dot)com> wrote:
>> For the backpatching, the patches sent previously here (=>
>>
>> http://www.postgresql.org/message-id/CAEB4t-OHNE95=n5U4ySsYkWipQsWeQuTBSJkaYJ63_1VzkzkhA@mail.gmail.com)
>> are fine IMO. They simply consist of a copy of what is done in
>> initdb.c. Now, perhaps we had better apply the patch duplicating the
>> logic to all branches, including HEAD, first, see what the buildfarm
>> says, and then finish wrapping up the refactoring patch.
>
> PFA patch for older branches, v1 patch was not applying cleaning other than
> 94. Thanks.

Cool, thanks for all the patches!
Nothing to say about the patch on master, except this thing not fixed:
#include "common/username.h"
+#include "common/restricted_token.h"
restricted_token.h should be declared before username.h.

In the back-branch patches, the format of the stderr messages is not
correct, and the suffix using progname is missing in pg_upgrade. For
example, this thing:
+ fprintf(stderr, _("could not open process token: error
code %lu\n"), GetLastError());
Should be rewritten like that for consistency with master and the
other utilities:
fprintf(stderr, _("%s: could not open process token: error code
%lu\n"), progname, GetLastError());
Perhaps now Bruce has a different opinion on the matter for pg_upgrade...

The 9.0 patch is missing as well, but I guess that the committer who
will pick up this series could get something up using the 9.1 patch...

I have also done some sanity checks down to 9.2 (got some issues when
building with ~9.1), and noticed no problems.
Regards,
--
Michael