[postgis-users] where to report bugs?

[Jaime Casanova]

On Sat, 6 Jun 2020 at 11:34, Darafei "Komяpa" Praliaskouski
<me at komzpa.net> wrote:
>
> Hello,
>
> You followed the best practice to report it. Trac is officially a bug tracker.
>
> Best way to get it fixed is to stomp it into a Github pull request with minimized version of your case in regress suite.
> Somewhere in here: https://github.com/postgis/postgis/tree/master/raster/test/regress - press "Edit" on a file and post a branch with that issue.
> After that, commit it as a pull request. CI (Travis) will run it and produce a non-optimized backtrace. It usually helps a tiny bit more, letting you understand what went wrong through all the "optimized out"-s.
>

interesting, you mean in addition to add it to trac? (i have received
the mantra now)

> Reasons why your report is hanging for now are:
>  - people have lives and postgis is a hobby;

maybe i'm just used to the response times in postgresql lists

>  - nobody really dug deep into raster codebase in last years, so you have a perfect opportunity to become world's best expert;

sadly, i don't actually know to much about GIS. i'm doing this
exercise by my own because some customers uses postgis and i prefer a
system that doesn't crash in simple ways

>  - it's synthetic query by a tool designed to break stuff, meaning there's no human being that wrote that query who you can empathize with to get thing fixed.
>

yes, you're right about the query being synthetic and sqlsmith being
designed to break things in odd ways.

having said that, this normally results in finding spots that could
result in hard to find errors.
this example could have bite people that have nulls in a raster column.

> I've looked at your query and it looks like the whole report should have been just one line:
>
> select st_union(null::raster, 4);
> server closed the connection unexpectedly
>        This probably means the server terminated abnormally
>        before or while processing the request.
>

curious, i did try to simplify it more but when i removed the
tablesample it didn't crash so i stoped there... but probably i had
already made the change to check the null...

in my defense it was too late for me at the time ;)


-- 
Jaime Casanova                      www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services