Re: libpq debug log

On 2021-Apr-02, Tom Lane wrote:

> On third thought, maybe we should push your patch too. Although I think
> 53aafdb9f is going to fix the platform-specific aspect of this, we are
> still going to risk some implementation dependence of the libpq_pipeline
> results:
>
> * Every so often, the number of digits in the reported line numbers
> will change (999 -> 1001 or the like), due to changes in unrelated
> code.

Bah -- of course.

> * Occasionally we refactor things so that the "same" error is reported
> from a different file.

True. (This was the reason for masking F and R, but I didn't realize
that it'd have an effect in the message length).

> It's hard to judge whether that will happen often enough to be an
> annoying maintenance problem, but there's definitely a hazard.
> Not sure if we should proactively lobotomize the test, or wait to
> see if we get annoyed.

I suspect we're going to see enough bf failures if we don't suppress it,
because the problem is going to only show up with TAP testing enabled
and the src/test/modules tests, which perhaps not all committers run.

> In any case I'd like to wait till after drongo's next run, so
> we can confirm whether or not the backslashes-in-__FILE__ hypothesis
> is correct. If it is, then 53aafdb9f is a good fix on its own
> merits, independently of libpq_pipeline.

Wilco.

--
Álvaro Herrera Valdivia, Chile
Essentially, you're proposing Kevlar shoes as a solution for the problem
that you want to walk around carrying a loaded gun aimed at your foot.
(Tom Lane)