berries garden tests failing

[robe]

Both berrie and berrie64 in 3.1 tests

16:31:08 COMMIT
16:31:08 -------------------------------------------------
16:31:08 Regression tests in progress (it will take time)
16:31:08 Result output: ./regress/postgis_garden_result.txt
16:31:08 -------------------------------------------------
16:31:08 psql --no-psqlrc -ad postgis_garden < ../doc/postgis_gardentest_31.sql > postgis_garden_result.txt 2>&1
16:57:31 make[1]: *** [Makefile:100: garden] Error 2
16:57:31 make[1]: Leaving directory '/home/jenkins/workspace/PostGIS_Worker_Run/label/berrie64/69b60e63f819ec26690e9b8cfcbe505904d62eab/regress'
16:57:31 make: *** [GNUmakefile:115: garden] Error 2
16:57:31 + err_status=2
16:57:31 + [ -d /home/jenkins/workspace/pg/label/berrie64/rel/pg14w64/data_55432/postmaster.pid ]
16:57:31 + exit 2
16:57:31 Build step 'Execute shell' marked build as failure
16:57:37 Finished: FAILURE

Seem to be okay in 3.2 and master branches. It's possible it's left over from me accidentally adding newer garden test cases we can't backport to 3.1

[robe]

Okay looks like it's crashing on:

SELECT ST_Node(ST_GeomFromText('LINESTRING EMPTY',4326))  As result

This is running with GEOS 3.10.2. The other branches where this is not crashing are running newer GEOS. I'm going to move berries PostGIS 3.1 runs to check this out with the latest released 3.10 to see if is still an issue.

Unfortunately I don't have that version of geos installed, so can't confirm it crashes on other than ARM.

[robe]

Switching the berries to GEOS 3.10.3 at [39ce9ea3/git] fixed this issue. So I guess it went undetected before because I recently made the garden script for 3.1 - 3.3 smarter about checking empties.