Opened 5 months ago
Last modified 2 days ago
#5771 new defect
address_standardizer_data_us upgrade sql script can be incorrect
Reported by: | evsi | Owned by: | robe |
---|---|---|---|
Priority: | medium | Milestone: | PostGIS 3.5.3 |
Component: | postgis | Version: | 3.4.x |
Keywords: | Cc: |
Description
Hi, there seems to be a problem with the way the way that address_standardizer_data_us upgrade sql script is handled during the build step in 3.4.0 and up. The address_standardizer_data_us upgrade file can end up being identical to that of address_standardizer.
To re-produce (I was using Amazon Linux 2 x86 GNU Make 3.82):
cd postgis-$POSTGIS_VERSION_NUM/extensions make touch ./address_standardizer/sql/address_standardizer_functions.sql make install diff address_standardizer/sql/address_standardizer--ANY--$POSTGIS_VERSION_NUM.sql address_standardizer/sql/address_standardizer_data_us--ANY--$POSTGIS_VERSION_NUM.sql | wc -l
and notice the result value is 0 (i.e. they are the exact same file), which is not expected.
(From my own looking)
When installation happens, the install-upgrade-paths target is called with EXTENSION set as address_standardizer_data_us link. This hits the generic target instead, which seems to be for address_standardizer link, rather than the one specifically for data_us.
When running make with parallelization the dependency 'sql/address_standardizer_functions.sql' can be newer than the target `sql/address_standardizer_data_us—ANY—<version>.sql', so the latter gets replaced with wrong upgrade script. (Hence the reproduction example just touches the file after make, to force the install target to "update" the script.)
Change History (4)
comment:1 by , 5 months ago
Milestone: | PostGIS 3.4.3 → PostGIS 3.4.4 |
---|
comment:2 by , 4 months ago
Owner: | changed from | to
---|
comment:3 by , 4 weeks ago
Milestone: | PostGIS 3.4.4 → PostGIS 3.5.2 |
---|
comment:4 by , 2 days ago
Milestone: | PostGIS 3.5.2 → PostGIS 3.5.3 |
---|
Ticket retargeted after milestone closed
I haven't had a chance to check if this is still an issue.