Thread: Re: [COMMITTERS] pgsql: Cause TestLib.pm to define $windows_os in all branches.

On Tue, Oct 13, 2015 at 8:35 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Cause TestLib.pm to define $windows_os in all branches.
>
> Back-port of a part of commit 690ed2b76ab91eb79ea04ee2bfbdc8a2693f2a37 that
> I'd depended on without realizing that it was only added recently.  Since
> it seems entirely likely that other such tests will need to be back-patched
> in future, providing the flag seems like a better answer than just putting
> a test in-line.

Is it really worth it back-patching the portions with $windows_os to
back-branches? This indeed makes back-patch a bit easier but the tests
cannot run on Windows because 13d856e1 has not been added in PG <=
9.5.
-- 
Michael

Michael Paquier <michael.paquier@gmail.com> writes:
> On Tue, Oct 13, 2015 at 8:35 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Cause TestLib.pm to define $windows_os in all branches.
>> 
>> Back-port of a part of commit 690ed2b76ab91eb79ea04ee2bfbdc8a2693f2a37 that
>> I'd depended on without realizing that it was only added recently.  Since
>> it seems entirely likely that other such tests will need to be back-patched
>> in future, providing the flag seems like a better answer than just putting
>> a test in-line.

> Is it really worth it back-patching the portions with $windows_os to
> back-branches? This indeed makes back-patch a bit easier but the tests
> cannot run on Windows because 13d856e1 has not been added in PG <=
> 9.5.

Hmm, well, why wasn't that back-patched?  We expect these tests to run
on Windows don't we?
        regards, tom lane

On Tue, Oct 13, 2015 at 11:31 AM, Tom Lane wrote:
> Hmm, well, why wasn't that back-patched?  We expect these tests to run
> on Windows don't we?

The message related to this particular commit is here:
http://www.postgresql.org/message-id/55B90161.5090506@iki.fi
I recall that we discussed about back-patching more such things to
improve the buildfarm coverage but I guess it fell from other's radar.
Would you consider pushing any sync-up patch for 9.5 and 9.4 I could
send? At quick glance, I think that's basically a combination of
adb4950, 13d856e1, 690ed2b and ff85fc8. Andrew, Noah, Heikki, and
others feel free to object of course if you think that's an utterly
bad idea.
-- 
Michael

On 10/12/2015 10:45 PM, Michael Paquier wrote:
> On Tue, Oct 13, 2015 at 11:31 AM, Tom Lane wrote:
>> Hmm, well, why wasn't that back-patched?  We expect these tests to run
>> on Windows don't we?
> The message related to this particular commit is here:
> http://www.postgresql.org/message-id/55B90161.5090506@iki.fi
> I recall that we discussed about back-patching more such things to
> improve the buildfarm coverage but I guess it fell from other's radar.
> Would you consider pushing any sync-up patch for 9.5 and 9.4 I could
> send? At quick glance, I think that's basically a combination of
> adb4950, 13d856e1, 690ed2b and ff85fc8. Andrew, Noah, Heikki, and
> others feel free to object of course if you think that's an utterly
> bad idea.


In general I think we can be a good deal more liberal about backpatching 
the testing regime than we are with production code, where we are always 
cautious, and the caution has paid big dividends in our reputation for 
stability.

cheers

andrew

On Tue, Oct 13, 2015 at 10:17 PM, Andrew Dunstan wrote:
> In general I think we can be a good deal more liberal about backpatching the
> testing regime than we are with production code, where we are always
> cautious, and the caution has paid big dividends in our reputation for
> stability.

Attached are patches for 9.4 and 9.5 syncing up everything with
master. I tested both of them on OSX, Linux and Windows (MSVC only
though using vcregress tapcheck).
--
Michael

Michael Paquier <michael.paquier@gmail.com> writes:
> On Tue, Oct 13, 2015 at 10:17 PM, Andrew Dunstan wrote:
>> In general I think we can be a good deal more liberal about backpatching the
>> testing regime than we are with production code, where we are always
>> cautious, and the caution has paid big dividends in our reputation for
>> stability.

> Attached are patches for 9.4 and 9.5 syncing up everything with
> master. I tested both of them on OSX, Linux and Windows (MSVC only
> though using vcregress tapcheck).

I'm not in a position to double-check that these patches work on Windows,
but I reviewed them through the expedient of diff'ing the patched files
against HEAD.  The only problem I found was you'd left out the
documentation addition about needing IPC::Run in install-windows.sgml.
So I've pushed them with that fix.

Where do we now stand on invoking the TAP tests in the buildfarm?
        regards, tom lane

On 11/17/2015 02:15 PM, Tom Lane wrote:
> Michael Paquier <michael.paquier@gmail.com> writes:
>> On Tue, Oct 13, 2015 at 10:17 PM, Andrew Dunstan wrote:
>>> In general I think we can be a good deal more liberal about backpatching the
>>> testing regime than we are with production code, where we are always
>>> cautious, and the caution has paid big dividends in our reputation for
>>> stability.
>> Attached are patches for 9.4 and 9.5 syncing up everything with
>> master. I tested both of them on OSX, Linux and Windows (MSVC only
>> though using vcregress tapcheck).
> I'm not in a position to double-check that these patches work on Windows,
> but I reviewed them through the expedient of diff'ing the patched files
> against HEAD.  The only problem I found was you'd left out the
> documentation addition about needing IPC::Run in install-windows.sgml.
> So I've pushed them with that fix.
>
> Where do we now stand on invoking the TAP tests in the buildfarm?
>
>             


This has fallen off my pile a bit :-( I will try to take a good look 
over the next 48 hours.

cheers

andrew

On Wed, Nov 18, 2015 at 4:15 AM, Tom Lane wrote:
> I'm not in a position to double-check that these patches work on Windows,
> but I reviewed them through the expedient of diff'ing the patched files
> against HEAD.

I'll double-check things a bit and post replies on this thread should
I detect a problem.

> The only problem I found was you'd left out the
> documentation addition about needing IPC::Run in install-windows.sgml.
> So I've pushed them with that fix.

Thanks, I clearly forgot this documentation bit.
-- 
Michael