Hi Tatsuo,
Thanks for rebasing to v43!
I noticed the Cirrus CI 32-bit build (test_world_32) failed on three
tests, all due to the same root cause in rpr_explain.out:
Sort Method: quicksort Memory: 27kB (expected, 64-bit)
Sort Method: quicksort Memory: 22kB (actual, 32-bit)
The rpr_explain_filter() function was already normalizing the RPR
Storage memory values, but missed the Sort node's "Memory: NNkB"
which also varies between 32-bit and 64-bit due to pointer/struct
size differences.
The attached incremental patch adds Sort Method memory normalization
to rpr_explain_filter(), using the same approach as the existing
Storage and Maximum Storage filters.
> The node serialization functions (141 lines, 0% coverage) are the
> > largest untested area. I'm not sure how to trigger these paths
> > in the regression test framework. Any suggestions?
>
> I think we can leave it as it is, until reluctant quantifier is
> implemented.
>
Agreed.
> > I'll send a separate email within a few days listing the FIXME
> > issues and other unresolved items from the mailing list discussion
> > for your review.
>
> Looking forward to reading your email.
Best regards,
Henson
diff --git a/src/test/regress/expected/rpr_explain.out
b/src/test/regress/expected/rpr_explain.out
index f23d06f6d59..bfe4ee2ba7d 100644
--- a/src/test/regress/expected/rpr_explain.out
+++ b/src/test/regress/expected/rpr_explain.out
@@ -33,7 +33,7 @@
-- DEFINE Expression Variations
-- Large Scale Statistics Verification
-- ============================================================
--- Filter function to normalize Storage memory values only (not NFA
statistics).
+-- Filter function to normalize platform-dependent memory values (not NFA
statistics).
-- NFA statistics should not change between platforms; if they do, it could
-- indicate issues such as uninitialized memory access.
-- Works for text, JSON, and XML formats.
@@ -45,7 +45,7 @@ declare
begin
for ln in execute $1
loop
- -- Normalize memory size in Storage line only (platform-dependent)
+ -- Normalize platform-dependent memory values
-- Keep NFA statistics numbers unchanged (they are test assertions)
-- Text format: "Storage: Memory Maximum Storage: 18kB"
@@ -63,6 +63,11 @@ begin
ln := regexp_replace(ln, '<Maximum-Storage>\d+</Maximum-Storage>',
'<Maximum-Storage>0</Maximum-Storage>', 'g');
end if;
+ -- Sort Method memory is platform-dependent (32-bit vs 64-bit)
+ if ln ~ 'Sort Method:.*Memory:' then
+ ln := regexp_replace(ln, 'Memory: \d+kB', 'Memory: NkB');
+ end if;
+
return next ln;
end loop;
end;
@@ -1625,7 +1630,7 @@ WINDOW w AS (
NFA: 54 absorbed (len 1/1/1.0), 18 skipped (len 1/1/1.0)
-> Sort (actual rows=90.00 loops=1)
Sort Key: p.p
- Sort Method: quicksort Memory: 27kB
+ Sort Method: quicksort Memory: NkB
-> Nested Loop (actual rows=90.00 loops=1)
-> Function Scan on generate_series p (actual rows=3.00
loops=1)
-> Function Scan on generate_series v (actual rows=30.00
loops=3)
@@ -1682,7 +1687,7 @@ WINDOW w AS (
NFA: 19 absorbed (len 1/1/1.0), 5 skipped (len 1/1/1.0)
-> Sort (actual rows=50.00 loops=1)
Sort Key: (CASE WHEN (v.v <= 25) THEN 1 ELSE 2 END)
- Sort Method: quicksort Memory: 26kB
+ Sort Method: quicksort Memory: NkB
-> Function Scan on generate_series v (actual rows=50.00 loops=1)
(12 rows)
diff --git a/src/test/regress/sql/rpr_explain.sql
b/src/test/regress/sql/rpr_explain.sql
index f8c8f62e594..ea8c39b991e 100644
--- a/src/test/regress/sql/rpr_explain.sql
+++ b/src/test/regress/sql/rpr_explain.sql
@@ -34,7 +34,7 @@
-- Large Scale Statistics Verification
-- ============================================================
--- Filter function to normalize Storage memory values only (not NFA
statistics).
+-- Filter function to normalize platform-dependent memory values (not NFA
statistics).
-- NFA statistics should not change between platforms; if they do, it could
-- indicate issues such as uninitialized memory access.
-- Works for text, JSON, and XML formats.
@@ -46,7 +46,7 @@ declare
begin
for ln in execute $1
loop
- -- Normalize memory size in Storage line only (platform-dependent)
+ -- Normalize platform-dependent memory values
-- Keep NFA statistics numbers unchanged (they are test assertions)
-- Text format: "Storage: Memory Maximum Storage: 18kB"
@@ -64,6 +64,11 @@ begin
ln := regexp_replace(ln, '<Maximum-Storage>\d+</Maximum-Storage>',
'<Maximum-Storage>0</Maximum-Storage>', 'g');
end if;
+ -- Sort Method memory is platform-dependent (32-bit vs 64-bit)
+ if ln ~ 'Sort Method:.*Memory:' then
+ ln := regexp_replace(ln, 'Memory: \d+kB', 'Memory: NkB');
+ end if;
+
return next ln;
end loop;
end;
