[postgis-devel] PSC Vote Request: I know it's early in 2.3 life cycle, but can we ditch PostgreSQL 9.1 in 2.3?

[Paragon Corporation]

Why do I want to ditch 9.1 in 2.3?

1) In 2.3, were are going to have to support 9.6, 9.5, 9.4, 9.3, 9.2
(actually I'd like to ditch 9.2 too but two in a row seems harsh).  That's 5
versions mind you (and 9.1 would make that 6), which means we probably have
lots of crap lying around handling old postgres apis instead of using newer
apis, just to compensate for supporting older version.  I think the real
droppers come at 9.2, but 2 in a row might be too much for some folks.

2) GUC junk has changed overtime.  In fact I think pramsey had to copy in
some GUC logic from newer PostgreSQL to compensate for the fact that piece
he needed is not exposed until later versions of PostgreSQL (I forget if the
new functionality was in 9.2 or 9.3) -- Paul you remember?  As I recall it
was 9.3 that the change happened but maybe other changes happened in 9.2 or
9.1
3) Side note -- I guess I got to figure this out -- but the new matrix  job
I setup on Debbie for PostGIS trunk to test 9.6,9.5, 9.4, 9.3, 9.2, 9.1 (
does not like 9.1)
Fails on the backend problem.  

PostgreSQL version:   PostgreSQL 9.1.19 -
http://debbie.postgis.net:8080/job/PostGIS_Regress/2343/consoleFull


sfcgal/empty .. failed (diff expected obtained:
/var/lib/jenkins/workspace/postgis/tmp/2_3_pg9.1w64/test_107_diff)
----------------------------------------------------------------------------
-
--- sfcgal/empty_expected	2015-11-28 06:43:43.982199307 +0000
+++ /var/lib/jenkins/workspace/postgis/tmp/2_3_pg9.1w64/test_107_out
2015-11-28 06:46:16.046248495 +0000
@@ -1,3 +1,4 @@
+ERROR:  unrecognized configuration parameter "postgis.backend"
 T1.1|POINT EMPTY
 T1.2|LINESTRING EMPTY
 T1.3|SRID=4326;POLYGON EMPTY
@@ -30,7 +31,7 @@
 ST_Intersection(empty, empty) == empty|t
 ST_Difference(geometry, empty) == geometry|t
 ST_Difference(empty, geometry) == empty|t
-ST_Distance(geometry, empty) == NULL|inf
+ST_Distance(geometry, empty) == NULL|
 ST_DWithin(geometry, empty, tolerance) == FALSE|f
 ST_Within(geometry, empty) == FALSE|f
 ST_Contains(empty, geometry) == FALSE|f
----------------------------------------------------------------------------
-
 sfcgal/geography .. failed (diff expected obtained:
/var/lib/jenkins/workspace/postgis/tmp/2_3_pg9.1w64/test_108_diff)


What's troubling me is why it works on the old non-matrix job unless I had
that setup wrong (like it wasn't testing sfcgal for 9.1) and since I have
standardized everything else to be the same except for pg version the issue
shows.

Also seems to work on 9.6,9.5, 9.4,9.3 and 9.2  so took out 9.1 for now
until I can isolate the issue or we decide to ditch 9.1 in PostGIS 2.3

Thanks,
Regina