Re: Still another race condition in recovery TAP tests

On Tue, Oct 17, 2017 at 5:01 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Fri, Oct 6, 2017 at 5:57 AM, Craig Ringer <craig(at)2ndquadrant(dot)com> wrote:
>>> Fewer people will test as we grow the list of modules they must first install.
>> At worst, all we have to do is provide a script
>> that fetches them, from distro repos if possible, and failing that
>> from CPAN.
>>
>> With cpanminus, that's pretty darn simple too.
>
> I think the idea that this is going to work for everybody is not very
> realistic. I am not going to run some random script to install stuff
> on my system in whatever manner it feels like, because when I do stuff
> like that I usually end up regretting it. If it's a throw-away VM,
> sure, but not otherwise.

Yeah, having something which is network-dependent to run the full set
of regression tests is not a good idea. I have seen users, at least in
Japan, deploying Postgres on hosts that have no external connection,
just a base OS image built with all packages present. Data center
rules can be very strict.

> We don't even have good documentation of what packages you should
> install on various packaging systems to build the server, let alone
> all of the additional things that may be required depending on build
> options and TAP tests you might want to run.

On top of that, let's not forget that we take the time to improve the
modules what we already have in the code tree. Let's not forget that
;)
--
Michael