From: | Nikolay Shaplov <dhyan(at)nataraj(dot)su> |
---|---|
To: | pgsql-hackers(at)postgresql(dot)org |
Cc: | Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>, Andres Freund <andres(at)anarazel(dot)de> |
Subject: | [PATCH] Tests for reloptions |
Date: | 2017-09-05 09:41:34 |
Message-ID: | 2615372.orqtEn8VGB@x200m |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
This thread continues discussion at https://www.postgresql.org/message-id/20170903094543.kkqdbdjuxwa5z6ji@alvherre.pgsql
(Shortly: I refactored reloptions code, Alvaro offered to commit tests before
the full patch)
> I see that this patch adds a few tests for reloptions, for instance in
> bloom. I think we should split this in at least two commits, one that
> adds those tests before the functionality change, so that they can be
> committed in advance, verify that the buildfarm likes it with the
> current code, and verify the coverage.
This sounds as a really good idea.
Though I have several questions. This tests also covers some functionality
that were introduced only in my patch:
1. Forbid SET and RESET options where they should not be changed
2. Forbid creating tables with toasts options when no toast table is created
3. Split StdRdOptions into HeapOptions and ToastOptions and forbid uising Heap
specific options for toast.
In the patch I've attached I've commented out this functionality. But I am not
quite sure that it is good idea to commit it this way in master.
May be it would be good to make 1-3 as a separate patches and bring it's tests
with, as a separate commit. But...
2nd can be easily ported to master. It does not depend much on my reloptions
patch as far as I remember.
It would be insane to port 1st feature to master. It highly integrated with
reloptions mechanism, It would require complete reimplementation of this
feature using old reloptions tools. I totally do not want to do it
The 3rd functionality came from philosophy one relation-type -- one options
catalog, that was implemented in my reloptions patch. It is not really needed
in master without the full patch. With some efforts I think it can be made as a
separate patch, thought I would also try to avoid it if possible.
So the questions still is: should we commit not existent functionality tests
commented, uncomented but with no proper error response in expected output, or
just remove these tests from this patch?
--
Do code for fun. Can do it for money (Perl & C/C++ ~10h/week)
Attachment | Content-Type | Size |
---|---|---|
relopt_tests_v1a.diff | text/x-patch | 32.3 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Amit Kapila | 2017-09-05 09:52:41 | Re: WIP: long transactions on hot standby feedback replica / proof of concept |
Previous Message | Chris Travers | 2017-09-05 09:31:07 | Re: Proposal: pg_rewind to skip config files |