Re: [PATCH] Add regression tests of ecpg command notice (error / warning)

On 2025/02/06 8:57, Ryo Kanbayashi wrote:
> On Wed, Feb 5, 2025 at 9:31 PM Ryo Kanbayashi <kanbayashi(dot)dev(at)gmail(dot)com> wrote:
>>
>> Hi hackers,
>>
>> When I wrote patch of ecpg command notice bug, I recognized needs of
>> regression tests for ecpg command notices and I say that I write the
>> tests.

Thanks for working on this!

>> I explain about implementation of this patch.
>>
>> What is this patch
>> - add regression tests which test ecpg command notices such as warning
>> and errors
>> - test notices implemented in ecpg.addons file
>>
>> Basic policy on implementation
>> - do in a way that matches the method using the existing pg_regress
>> command as much as possible
>> - avoid methods that increase the scope of influence
>>
>> Next, I list answers to points that are likely to be pointed out in
>> advance below :)
>> - shell scripts and bat files is used due to ...
>> avoid non zero exit code of ecpg command makes tests failure
>> avoid increasing C code for executing binary which cares cross platform
>> - python code is used because I couldn't write meson.build
>> appropriately describe dependency about materials which is used on
>> tests without it. please help me...
>> - as you said, kick this kind of tests by pg_regress accompanied with
>> needless PG server process execution. but pg_regress doesn't execute
>> test without it and making pg_regress require modification which has
>> not small scope of influence

Wouldn't it be simpler to use the existing TAP test mechanism,
as shown in the attached patch? Please note that this patch is very WIP,
so there would be many places that need further implementation and refinement.

Regards,

--
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION