Horton

Thread

Commits

  1. Remove operator_precedence_warning.

  1. Removal of operator_precedence_warning

    Tom Lane <tgl@sss.pgh.pa.us> — 2020-12-04T21:39:05Z

    I think it's time for $SUBJECT.  We added this GUC in 9.5, which
will be EOL by the time of our next major release, and it was never
meant as more than a transitional aid.  Moreover, it's been buggy
as heck (cf abb164655, 05104f693, 01e0cbc4f, 4cae471d1), and the
fact that some of those errors went undetected for years shows that
it's not really gotten much field usage.

Hence, I propose the attached.  Comments?

			regards, tom lane

  2. Re: Removal of operator_precedence_warning

    Alvaro Herrera <alvherre@alvh.no-ip.org> — 2020-12-04T22:01:38Z

    On 2020-Dec-04, Tom Lane wrote:

> I think it's time for $SUBJECT.  We added this GUC in 9.5, which
> will be EOL by the time of our next major release, and it was never
> meant as more than a transitional aid.  Moreover, it's been buggy
> as heck (cf abb164655, 05104f693, 01e0cbc4f, 4cae471d1), and the
> fact that some of those errors went undetected for years shows that
> it's not really gotten much field usage.
> 
> Hence, I propose the attached.  Comments?

I wonder if it'd be fruitful to ask the submitters of those bugs about
their experiences with the feature.  Did they find it useful in finding
precedence problems in their code?  Did they experience other problems
that they didn't report?

Reading the reports mentioned in those commits, it doesn't look like any
of them were actually using the feature -- they all seem to have come
across the problems by accidents of varying nature.

  3. Re: Removal of operator_precedence_warning

    Tom Lane <tgl@sss.pgh.pa.us> — 2020-12-04T22:21:12Z

    Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> Reading the reports mentioned in those commits, it doesn't look like any
> of them were actually using the feature -- they all seem to have come
> across the problems by accidents of varying nature.

The two oldest reports look like the submitters had
operator_precedence_warning turned on in normal use, which is reasonable
given that was early 9.5.x days.  The third one looks like it was a test
setup, while the latest bug sounds like it was found by code inspection
not by stumbling over the misbehavior.  So people did use it, at least
for awhile.  But anyone who's going directly from 9.4 or earlier to v14
is going to have lots more compatibility issues to worry about besides
precedence.

			regards, tom lane