Random errors on GHA postgis install some sort of incomplete scripts

[robe]

I just noticed GHA has been failing for a couple of runs, but it's random.

So guessing it has to do with parallel jobs?

​https://github.com/postgis/postgis/actions/runs/10935290396/job/30356784011

Errors like

_postgis_drop_function_by_signature 
-------------------------------------
 
(1 row)

psql:/src/postgis/regress/00-regress-install/share/contrib/postgis/rtpostgis_upgrade.sql:843: ERROR:  syntax error at or near "DECLARE"
LINE 122: DECLARE

and

 
(1 row)

psql:/src/postgis/regress/00-regress-install/share/contrib/postgis/rtpostgis_upgrade.sql:4433: ERROR:  syntax error at or near "CREATE"
LINE 39: CREATE OR REPLACE FUNCTION _postgis_drop_function_by_identit...
         ^
CONTEXT:  invalid type name "CREATE OR REPLACE FUNCTION _postgis_drop_function_by_identity(
	function_name text,
	function_arguments text,
	deprecated_in_version text "

I know strk had completed some changes recently for parallelism, which is around the time these failures started happening.

The bots failing is random. It's not all but across various PG versions, GEOS versions, and GDAL versions

[robe]

Just a note about the commonality of these, they all involve random syntax errors in raster scripts primarily the rtpostgis_upgrade.sql script

and here is another pattern of failure not with sql script errors

#strip BEGIN/COMMIT since these are not allowed in extensions
make[3]: Leaving directory '/src/postgis/raster/rt_pg'
make[3]: *** [Makefile:156: rtpostgis_upgrade.sql.in] Error 255
make[3]: *** Deleting file 'rtpostgis_upgrade.sql.in'
make[3]: *** Waiting for unfinished jobs....
make[2]: *** [Makefile:76: ../../raster/rt_pg/rtpostgis_upgrade_cleanup.sql] Error 2
make[2]: *** Deleting file '../../raster/rt_pg/rtpostgis_upgrade_cleanup.sql'
make[3]: *** [Makefile:156: rtpostgis_upgrade.sql.in] Error 255
make[2]: *** Waiting for unfinished jobs....
make[3]: *** Waiting for unfinished jobs....
# WARNING: we should make sure these are not internal to function bodies
cat ../../postgis/common_before_upgrade.sql ../../postgis/postgis_before_upgrade.sql sql/postgis_upgrade_for_extension.sql.in ../../postgis/postgis_after_upgrade.sql ../../postgis/common_after_upgrade.sql | /usr/bin/perl -pe 's/BEGIN\;//g ; s/COMMIT\;//g' > sql/postgis_upgrade_for_extension.sql
make[2]: *** [Makefile:76: ../../raster/rt_pg/uninstall_rtpostgis.sql] Error 2
make[3]: Leaving directory '/src/postgis/raster/rt_pg'
make[2]: *** Deleting file '../../raster/rt_pg/uninstall_rtpostgis.sql'
make[3]: Leaving directory '/src/postgis/raster/rt_pg'
make[2]: *** [Makefile:76: ../../raster/rt_pg/rtpostgis_drop.sql] Error 2
make[2]: Leaving directory '/src/postgis/extensions/postgis'
make[2]: *** Deleting file '../../raster/rt_pg/rtpostgis_drop.sql'
make[1]: *** [Makefile:36: all] Error 1
make[1]: Leaving directory '/src/postgis/extensions'
make: *** [GNUmakefile:34: all] Error 1

[robe]

Might be a duplicate of #5779 but this issue started happening after recent commits for #5779 so assuming it was more a result of the efforts trying to fix that one.

I should note the GHA jobs are building with

make -j

So perhaps they use the max processes available or a random set

[strk]

Temptative fix in [be1c2273cdfe0490a43b0148bb8f18ce57a63916/git]

[robe]

Looks like your fix fixed it, so closing this one out.