Paper over regression failures in infinite_recurse() on PPC64 Linux.

Our infinite_recurse() test to verify sane stack-overrun behavior
is affected by a bug of the Linux kernel on PPC64: it will get SIGSEGV
if it receives a signal when the stack depth is (a) over 1MB and
(b) within a few kB of filling the current physical stack allocation.
See https://bugzilla.kernel.org/show_bug.cgi?id=205183.

Since this test is a bit time-consuming and we run it in parallel with
test scripts that do a lot of DDL, it can be expected to get an sinval
catchup interrupt at some point, leading to failure if the timing is
wrong.  This has caused more than 100 buildfarm failures over the
past year or so.

While a fix exists for the kernel bug, it might be years before that
propagates into all production kernels, particularly in some of the
older distros we have in the buildfarm.  For now, let's just back off
and not run this test on Linux PPC64; that loses nothing in test
coverage so far as our own code is concerned.

To do that, split this test into a new script infinite_recurse.sql
and skip the test when the platform name is powerpc64...-linux-gnu.

Back-patch to v12.  Branches before that have not been seen to get
this failure.  No doubt that's because the "errors" test was not
run in parallel with other tests before commit 798070ec0, greatly
reducing the odds of an sinval catchup being necessary.

I also back-patched 3c8553547 into v12, just so the new regression
script would look the same in all branches having it.

Discussion: https://postgr.es/m/3479046.1602607848@sss.pgh.pa.us
Discussion: https://postgr.es/m/20190723162703.GM22387%40telsasoft.com

7 files changed, 71 insertions, 20 deletions

diff --git a/src/test/regress/expected/errors.out b/src/test/regress/expected/errors.out
index a525aa2f937..1e7b5a70461 100644
--- a/src/test/regress/expected/errors.out
+++ b/src/test/regress/expected/errors.out
@@ -440,13 +440,3 @@ NULL);
 ERROR:  syntax error at or near "NUL"
 LINE 16: ...L, id2 TEXT NOT NULL PRIMARY KEY, id3 INTEGER NOT NUL, id4 I...
                                                               ^
--- Check that stack depth detection mechanism works and
--- max_stack_depth is not set too high.  The full error report is not
--- very stable, so show only SQLSTATE and primary error message.
-create function infinite_recurse() returns int as
-'select infinite_recurse()' language sql;
-\set VERBOSITY sqlstate
-select infinite_recurse();
-ERROR:  54001
-\echo :LAST_ERROR_MESSAGE
-stack depth limit exceeded
diff --git a/src/test/regress/expected/infinite_recurse.out b/src/test/regress/expected/infinite_recurse.out
new file mode 100644
index 00000000000..aa102fadd83
--- /dev/null
+++ b/src/test/regress/expected/infinite_recurse.out
@@ -0,0 +1,24 @@
+-- Check that stack depth detection mechanism works and
+-- max_stack_depth is not set too high.
+create function infinite_recurse() returns int as
+'select infinite_recurse()' language sql;
+-- Unfortunately, up till mid 2020 the Linux kernel had a bug in PPC64
+-- signal handling that would cause this test to crash if it happened
+-- to receive an sinval catchup interrupt while the stack is deep:
+-- https://bugzilla.kernel.org/show_bug.cgi?id=205183
+-- It is likely to be many years before that bug disappears from all
+-- production kernels, so disable this test on such platforms.
+-- (We still create the function, so as not to have a cross-platform
+-- difference in the end state of the regression database.)
+SELECT version() ~ 'powerpc64[^,]*-linux-gnu'
+       AS skip_test \gset
+\if :skip_test
+\quit
+\endif
+-- The full error report is not very stable, so we show only SQLSTATE
+-- and primary error message.
+\set VERBOSITY sqlstate
+select infinite_recurse();
+ERROR:  54001
+\echo :LAST_ERROR_MESSAGE
+stack depth limit exceeded
diff --git a/src/test/regress/expected/infinite_recurse_1.out b/src/test/regress/expected/infinite_recurse_1.out
new file mode 100644
index 00000000000..b2c99a0d0d4
--- /dev/null
+++ b/src/test/regress/expected/infinite_recurse_1.out
@@ -0,0 +1,16 @@
+-- Check that stack depth detection mechanism works and
+-- max_stack_depth is not set too high.
+create function infinite_recurse() returns int as
+'select infinite_recurse()' language sql;
+-- Unfortunately, up till mid 2020 the Linux kernel had a bug in PPC64
+-- signal handling that would cause this test to crash if it happened
+-- to receive an sinval catchup interrupt while the stack is deep:
+-- https://bugzilla.kernel.org/show_bug.cgi?id=205183
+-- It is likely to be many years before that bug disappears from all
+-- production kernels, so disable this test on such platforms.
+-- (We still create the function, so as not to have a cross-platform
+-- difference in the end state of the regression database.)
+SELECT version() ~ 'powerpc64[^,]*-linux-gnu'
+       AS skip_test \gset
+\if :skip_test
+\quit
diff --git a/src/test/regress/parallel_schedule b/src/test/regress/parallel_schedule
index 026ea880cde..ae89ed7f0b4 100644
--- a/src/test/regress/parallel_schedule
+++ b/src/test/regress/parallel_schedule
@@ -55,7 +55,7 @@ test: create_index create_index_spgist create_view index_including index_includi
 # ----------
 # Another group of parallel tests
 # ----------
-test: create_aggregate create_function_3 create_cast constraints triggers select inherit typed_table vacuum drop_if_exists updatable_views roleattributes create_am hash_func errors
+test: create_aggregate create_function_3 create_cast constraints triggers select inherit typed_table vacuum drop_if_exists updatable_views roleattributes create_am hash_func errors infinite_recurse
 
 # ----------
 # sanity_check does a vacuum, affecting the sort order of SELECT *
diff --git a/src/test/regress/serial_schedule b/src/test/regress/serial_schedule
index 979d9261197..525bdc804f6 100644
--- a/src/test/regress/serial_schedule
+++ b/src/test/regress/serial_schedule
@@ -83,6 +83,7 @@ test: roleattributes
 test: create_am
 test: hash_func
 test: errors
+test: infinite_recurse
 test: sanity_check
 test: select_into
 test: select_distinct
diff --git a/src/test/regress/sql/errors.sql b/src/test/regress/sql/errors.sql
index 86b672538a1..66a56b28f62 100644
--- a/src/test/regress/sql/errors.sql
+++ b/src/test/regress/sql/errors.sql
@@ -364,12 +364,3 @@ INT4
 UNIQUE
 NOT
 NULL);
-
--- Check that stack depth detection mechanism works and
--- max_stack_depth is not set too high.  The full error report is not
--- very stable, so show only SQLSTATE and primary error message.
-create function infinite_recurse() returns int as
-'select infinite_recurse()' language sql;
-\set VERBOSITY sqlstate
-select infinite_recurse();
-\echo :LAST_ERROR_MESSAGE
diff --git a/src/test/regress/sql/infinite_recurse.sql b/src/test/regress/sql/infinite_recurse.sql
new file mode 100644
index 00000000000..151dba4a7ae
--- /dev/null
+++ b/src/test/regress/sql/infinite_recurse.sql
@@ -0,0 +1,29 @@
+-- Check that stack depth detection mechanism works and
+-- max_stack_depth is not set too high.
+
+create function infinite_recurse() returns int as
+'select infinite_recurse()' language sql;
+
+-- Unfortunately, up till mid 2020 the Linux kernel had a bug in PPC64
+-- signal handling that would cause this test to crash if it happened
+-- to receive an sinval catchup interrupt while the stack is deep:
+-- https://bugzilla.kernel.org/show_bug.cgi?id=205183
+-- It is likely to be many years before that bug disappears from all
+-- production kernels, so disable this test on such platforms.
+-- (We still create the function, so as not to have a cross-platform
+-- difference in the end state of the regression database.)
+
+SELECT version() ~ 'powerpc64[^,]*-linux-gnu'
+       AS skip_test \gset
+\if :skip_test
+\quit
+\endif
+
+-- The full error report is not very stable, so we show only SQLSTATE
+-- and primary error message.
+
+\set VERBOSITY sqlstate
+
+select infinite_recurse();
+
+\echo :LAST_ERROR_MESSAGE