Re: Sigh, I broke crake again - Mailing list pgsql-hackers

From Andrew Dunstan
Subject Re: Sigh, I broke crake again
Date
Msg-id CAD5tBcLPFGEhuokgyW+SvbiT3O__Dy8Kkj7oa_o3P1P60h0DQg@mail.gmail.com
Whole thread Raw
In response to Re: Sigh, I broke crake again  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers


On Wed, Feb 28, 2018 at 6:08 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
I wrote:
> I had not intended to back-patch, since those changes were just cosmetic,
> but it might be the best way to preserve the XversionUpgrade tests.

After closer study, it seems like the least painful answer is:

1. In HEAD, revert the renaming of int44in/int44out to
city_budget_in/_out; the gain in obviousness isn't worth the screwing
around we'd have to do to cope with it in cross-version upgrade testing.

2. In the back branches, drop the CREATE FUNCTION commands for
funny_dup17 and boxarea, which are used nowhere so they can easily
be dispensed with.

We could ask TestUpgradeXversion.pm to drop the latter two functions,
similarly to what it's already doing for oldstyle_length.  But
oldstyle_length was actually being used for something in the older
branches so it was worth the trouble to cope with a cross-branch
difference.  These two don't seem worth changing the buildfarm for.

                  


I see you're reverted the change that caused the issue. The fact is that we're always going to get this sort of issue from time to time with cross-version testing. FWIW, my experience in testing the module offline over an extended period (several years) is that the breakage isn't very frequent. I'm always willing to make small tweaks to accommodate things like name changes. It's not terribly difficult. I think it's sufficiently important that we have good cross-version upgrade testing that we can tolerate a little temporary breakage occasionally.

cheers

andrew
 

pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: Reopen logfile on SIGHUP
Next
From: "David G. Johnston"
Date:
Subject: Re: Reopen logfile on SIGHUP