make garden fails

[strk]

I don't really know what's the error triggering the failure. Here's the full log:

[strk@cdb:/usr/src/postgis/postgis(svn-trunk)] make garden
------------------------------------------------------
Generating SQL file from Documentation
------------------------------------------------------
make -C doc garden
make[1]: Entering directory `/usr/src/postgis/postgis/doc'
/usr/bin/xsltproc -o postgis_gardentest_22.sql xsl/postgis_gardentest.sql.xsl postgis-out.xml
/usr/bin/xsltproc -o raster_gardentest_22.sql xsl/raster_gardentest.sql.xsl postgis-out.xml
make[1]: Leaving directory `/usr/src/postgis/postgis/doc'
------------------------------------------------------
Launch regression Test
------------------------------------------------------
make -C regress garden
make[1]: Entering directory `/usr/src/postgis/postgis/regress'
createdb postgis_garden
createlang plpgsql postgis_garden
createlang: language "plpgsql" is already installed in database "postgis_garden"
make[1]: *** [garden] Error 2
make[1]: Leaving directory `/usr/src/postgis/postgis/regress'
make: *** [garden] Error 2

[robe]

weird I don't even have a createlang plpgsql in the script. Olivier had set up the

make garden

Never looked to see what was behind it.

Does it generate the .sql files okay on your system?

[strk]

doc/postgis_gardentest_22.sql is generated, yes. I see the "garden" rule in regress/Makefile does a manual database creation and run of psql, and wondering if it would make sense to have that done by run_test.pl directly. What would the "expected" result be ? We'd need a way for run_test.pl to evaluate the results…

[strk]

r12424 fixes it. Very useful. Indeed it handled to crash the backend… Not sure how to now figure out which query did that. Robe ?

[strk]

Last lines of the log file (postgis_garden_result.txt):

                    ?column?                     
-------------------------------------------------
 GeometryST_AsTWKBAgg Single NULL: Start Testing
(1 row)

INSERT 0 1
BEGIN
connection to server was lost

[strk]

Here's the offending query:

SELECT ST_AsTWKBAgg(foo1.the_geom, 3) As result FROM ((SELECT CAST(Null As geometry) As the_geom)) As foo1 LIMIT 3;

\cc @nicklas

[robe]

strk,

Looks like you figured it out. Hmm that used to crash my backend too but last I tried it it didn't after nicklas put in some patches after I put in #2672 . You sure you are running latest 2.2?

BTW to figure out the crashing entry the script writes to tables in the database. The last log entry in the postgis_garden_log22 (or 21 or whatever) is the crashing query since it writes the test before it runs it. Output of tests go in postgis_garden_log22_output

[robe]

strk — I reopened #2672 for your comment and closing this one out.

[robe]

backported to 2.1 branch at r12444