[SCM] PostGIS branch stable-3.6 updated. 3.6.4-135-gbef1bf6b2
git at osgeo.org
git at osgeo.org
Sat Aug 15 09:36:44 PDT 2026
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "PostGIS".
The branch, stable-3.6 has been updated
via bef1bf6b2aa21a6e92af072aae8b19f39e6520e4 (commit)
via c82d1e42ff057036b597500fc89ee4925c98f71a (commit)
via e861a4d0e6acf4740a2489b28105a8565c1ffb71 (commit)
from 1eae54a867edd9457fdfe87c9cca2b0c28824a57 (commit)
Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.
- Log -----------------------------------------------------------------
commit bef1bf6b2aa21a6e92af072aae8b19f39e6520e4
Merge: 1eae54a86 c82d1e42f
Author: Darafei Praliaskouski <komzpa at gmail.com>
Date: Sat Aug 15 09:36:43 2026 -0700
Merge pull request '[3.6] Fix address_standardizer test-security_bounds installcheck failure' (!736) from Komzpa/postgis:fix/addrstd-security-bounds-3.6-20260815 into stable-3.6
The `test-security_bounds` regression test added by [PR 709](https://gitea.osgeo.org/postgis/postgis/pulls/709) failed `installcheck` on every platform that actually runs it, including a clean Linux build: pg_regress reported "1 of 6 tests failed" with a whitespace-only diff. Real psql right-pads an aligned-mode single-column header with a trailing space, the same convention already used by this extension's pre-existing `test-debug_standardize_address` fixture, and the added fixture was missing that space plus the trailing blank line pg_regress appends. All hardening assertions, including the 254/255/256/300-byte scanner-token boundary and oversized or invalid rule rejection, already matched exactly, so only the captured fixture is corrected.
This also extends `.editorconfig` so the PGXS `expected/*.out` fixture convention this extension uses gets the same trim-trailing-whitespace protection PostGIS's own `*_expected` fixtures already have, which is the gap that let the broken fixture ship.
Note: Woodpecker does not run `make -C extensions/address_standardizer installcheck` on any branch, so this failure was only visible on Jenkins Winnie; validated locally on Linux with a clean build (6 of 6 tests pass).
References https://gitea.osgeo.org/postgis/postgis/pulls/709
Reviewed-on: https://gitea.osgeo.org/postgis/postgis/pulls/736
commit c82d1e42ff057036b597500fc89ee4925c98f71a
Author: Darafei Praliaskouski <me at komzpa.net>
Date: Sat Aug 15 20:15:02 2026 +0400
Protect PGXS extension regression fixtures from whitespace trimming
.editorconfig's "*_expected" exception to trim_trailing_whitespace only
matches PostGIS's own regress/core/<test>_expected naming convention.
extensions/address_standardizer uses the PGXS pg_regress convention
instead, expected/<test>.out, which the glob does not match, so an
editor honoring .editorconfig silently strips the trailing whitespace
that real psql aligned-mode output requires in these files. That is
what produced the test-security_bounds fixture defect fixed in the
previous commit. Extend the exception to cover expected/*.out at any
depth so future PGXS-style fixtures do not repeat it.
diff --git a/.editorconfig b/.editorconfig
index d9aad2c76..7238d42e7 100644
--- a/.editorconfig
+++ b/.editorconfig
@@ -17,6 +17,13 @@ indent_size = 2
[*_expected]
trim_trailing_whitespace = false
+# PGXS-style extension regression fixtures (extensions/*/expected/*.out)
+# use pg_regress's own naming convention instead of *_expected, and need
+# the same protection: aligned-mode psql output legitimately ends lines
+# in trailing whitespace.
+[**/expected/*.out]
+trim_trailing_whitespace = false
+
# C files want tab indentation
[*.{c,h,h.in}]
indent_style = tab
commit e861a4d0e6acf4740a2489b28105a8565c1ffb71
Author: Darafei Praliaskouski <me at komzpa.net>
Date: Sat Aug 15 20:14:49 2026 +0400
Fix test-security_bounds expected output header padding
The new address_standardizer regression test added by the pre-split
address component hardening (GT-709) failed installcheck on every
platform, including a clean Linux build: pg_regress reported
"1 of 6 tests failed" for test-security_bounds.
The regression.diffs is whitespace-only. Real, unmodified psql
right-pads a single-column aligned-mode header to the border width
with a trailing space (compare the pre-existing, unrelated
expected/test-debug_standardize_address.out, whose
"jsonb_array_length " header already carries the same trailing
space); the data row below it does not need padding once its width
already equals the header width. The expected/test-security_bounds.out
fixture added by GT-709 has that trailing space missing from both of
its "*_ok" headers, and is missing the trailing blank line pg_regress
always appends after the last statement's output. Every substantive
assertion in the file (the NOTICE and error text, the accept/reject
behavior at the 254/255/256/300-byte token boundaries, the malformed
rule handling) is untouched and already matched byte for byte.
This is a fixture authoring defect, not a behavior regression: the
address_standardizer hardening itself works as intended. The
.editorconfig "*_expected" trim-trailing-whitespace exception in this
repository matches PostGIS's regress/core naming convention
(<test>_expected) but not the extensions/*/expected/<test>.out PGXS
convention that address_standardizer uses, so any editor honoring
.editorconfig strips this exact header padding on save; a following
commit closes that gap.
References https://gitea.osgeo.org/postgis/postgis/pulls/709
diff --git a/extensions/address_standardizer/expected/test-security_bounds.out b/extensions/address_standardizer/expected/test-security_bounds.out
index 8bfa88c8e..7261ba068 100644
--- a/extensions/address_standardizer/expected/test-security_bounds.out
+++ b/extensions/address_standardizer/expected/test-security_bounds.out
@@ -65,7 +65,7 @@ BEGIN
END
$scanner_bounds$;
SELECT 'scanner_bounds_ok' AS scanner_bounds_ok;
- scanner_bounds_ok
+ scanner_bounds_ok
-------------------
scanner_bounds_ok
(1 row)
@@ -116,7 +116,8 @@ $$;
NOTICE: load_roles: failed to add rule 1 (7): 1 -1 5 -1 5 0
NOTICE: bad-rule-type: CreateStd: failed to load 't_bad_rule_type' for rules
SELECT 'security_bounds_ok' AS security_bounds_ok;
- security_bounds_ok
+ security_bounds_ok
--------------------
security_bounds_ok
(1 row)
+
-----------------------------------------------------------------------
Summary of changes:
.editorconfig | 7 +++++++
extensions/address_standardizer/expected/test-security_bounds.out | 5 +++--
2 files changed, 10 insertions(+), 2 deletions(-)
hooks/post-receive
--
PostGIS
More information about the postgis-tickets
mailing list