"Fix all tickets" experiment interim report

[Darafei "Komяpa" Praliaskouski]

Hi,

 - we had ~331 open tickets in PostGIS, which we have previously classified
as not critical or into the Fund Me category.
https://trac.osgeo.org/postgis/report/31. These user requests have been
hanging around in the backlog for 5-15 years typically.

 - for at least around 200 (~80%) of these it is possible to produce a
patch and test coverage using AI (codex-cli). I've pushed them as pull
requests on github to get a review from CodeRabbit. I have not merged any
"blindly".

 - non-fixable tickets are due to various reasons - related to bug reports
to upstream GDAL, GEOS, SFCGAL or Postgres machinery; major design changes
(coordinate epochs, exact curve area, geography clustering ...), or working
in combination with something else (pgAdmin, OSGeoLive, windows installer,
etc); many topology tickets covering architecture that needs to be designed
(batch loading, repair utilities, index churn, staged noding) or similar
raster design gaps (masks, aggregate semantics, multi-raster GDAL export).

 - some of the tickets proved to be invalid and discussion in the chat led
us to understanding how to improve documentation (need better explanations
on why planar operations on geodetic crs aren't always giving expected
results).

 - some improvements in tests (new oss-fuzz targets) uncovered more issues
to get fixed in existing codebase - thanks Even Rouault for addressing some
of these before.

 - discussion on adopting AI/LLM policy is also going on. Something along
the lines of Linux Kernel's
https://docs.kernel.org/process/coding-assistants.html - generally covered
by existing policies, but with separate entry point for the interested in
it specifically.

 - 43 of the smaller PRs/tickets I had the time this week to review,
iterate on and get merged.

 - got some first thanks from users.

Thanks Sandro Santilli, Loïc Bartoletti, Lars Aksel Opsahl, Regina Obe,
Paul Ramsey, Dan Baston for bringing in your comments, reviews and notes
this week.

What's next:

 - A side product of this experiment is a set of notes on how the developer
documentation needs to be improved so that some of the "obvious" things and
procedures are recorded somewhere. I'll clean it up and bring it to a
separate review, I shared drafts in the matrix chat where myself and Sandro
are iterating on how to make docs target people and not machines.

 - Trac is showing its age. There is a slow ongoing discussion on how to
move to another forge, and how to not lose the tickets/id's/links in the
process. My hypothesis is that we can radically minimize the number of open
tickets before implementing any migration and thus have a much smoother
move. Reorganizing developer docs in repo to supersede trac wiki could also
help.

 - I plan to continue the review of the produced patches for closing the
currently open user's tickets.

Thanks,
Darafei.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20260622/820ac471/attachment.htm>