On 05.11.2015 13:49, Craig Ringer wrote:

I believe that we need to lower the barrier for testing.

While I agree, I'd also like to note that formulaic testing is often
of limited utility. Good testing still requires a significant
investment of time and effort to understand the changes made by a
patch, which areas need focused attention, think about corner cases,

Yes, you are right. But a limited test is better than no test at all. But of course not enough.

For me it is easy to check comments or sql commands, but not the c code. But with lower barriers it would be easier to test 2 of the 3 mentioned items. At the moment its often none, because its hard.

"make check passes" doesn't really tell anyone that much.

I could even imagine to set up an open for everyone test-instance of HEAD
where users are allowed to test like the wanted. Than the barrier is reduced
to "connect to PostgreSQL and execute SQL".

Gee, that'd be fun to host ;)
More seriously, it's not HEAD that's of that much interest, it's HEAD
+ [some patch or set of patches].

There are systems that can pull in patchsets, build a project, and run
it. But for something like PostgreSQL it'd be pretty hard to offer
wide public access, given the trivial potential for abuse.

Yes, but i would do this. Creating a FreeBSD Jail which is reset regularly is no great deal and very secure. My bigger problem is the lack of IPv4 addresses. At the moment i am limited to IPv6 only hosts.


Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:

Reply via email to