riscv64 regression test issues

[df7cb]

pljs finally got accepted into Debian main, exposing it to more architectures than on apt.postgresql.org.

riscv64 is raising these bits:

**** regression.diffs ****
diff -U3 /build/reproducible-path/pljs-1.0.3/expected/function.out /build/reproducible-path/pljs-1.0.3/results/function.out
--- /build/reproducible-path/pljs-1.0.3/expected/function.out	2025-08-21 05:05:15.000000000 +0000
+++ /build/reproducible-path/pljs-1.0.3/results/function.out	2025-10-14 10:56:48.632056150 +0000
@@ -198,7 +198,7 @@
 -------------
         -0.2
            1
-         0.2
+            
 (3 rows)
 
 CREATE FUNCTION set_of_unnamed_records() RETURNS SETOF record AS
@@ -211,7 +211,7 @@
 SELECT * FROM set_of_unnamed_records() t (i bool);
  i 
 ---
- t
+ 
 (1 row)
 
 CREATE OR REPLACE FUNCTION set_of_unnamed_records() RETURNS SETOF record AS
diff -U3 /build/reproducible-path/pljs-1.0.3/expected/array_spread.out /build/reproducible-path/pljs-1.0.3/results/array_spread.out
--- /build/reproducible-path/pljs-1.0.3/expected/array_spread.out	2025-08-21 05:05:15.000000000 +0000
+++ /build/reproducible-path/pljs-1.0.3/results/array_spread.out	2025-10-14 10:57:01.552014194 +0000
@@ -5,7 +5,7 @@
 [...({})];
 $$ language pljs;
 ERROR:  execution error
-DETAIL:  InternalError: out of memory
+DETAIL:  InternalError: interrupted
     at <anonymous> (<function>:2:1)
     at <eval> (<function>:3:3)
 
@@ -13,7 +13,7 @@
 [...({})];
 $$ language pljs;
 ERROR:  execution error
-DETAIL:  InternalError: out of memory
+DETAIL:  InternalError: interrupted
     at <anonymous> (<function>:2:1)
     at <eval> (<function>:3:3)

Full build log: https://buildd.debian.org/status/fetch.php?pkg=pljs&arch=riscv64&ver=1.0.3-1&stamp=1760439440&raw=0

Other issues (all of these are big-endian): s390x: build seems to time out ppc64 (non-el), sparc64: crash https://buildd.debian.org/status/package.php?p=pljs ... but these are not release architectures, so we don't really have to care if we don't want to

[JerrySievert]

thanks much @df7cb!

I will look to see if my only riscv (it's an extremely early sbc, and likely 32bit) exhibits the issue, but I will have to find which box it is hiding in. otherwise, I'll order a small sbc - it will be good to have one to validate on (I currently validate on arm64 and x86_64 under both linux and macOS).

[df7cb]

Actually, I was too quick in summarizing that, s390x is a release architecture. But let's tackle one problem after another.

[JerrySievert]

hopefully, solving one will solve all of them, but it will take diving in to know for sure

[df7cb]

riscv64 is happy now, thanks!

There are a few (big-endian?) problems left on https://buildd.debian.org/status/package.php?p=pljs, though.

[JerrySievert]

array_spread in this (and the risc-v case previously) are due to the amount of time it takes for the test to fail. I had upped that significantly, which allowed for the test to pass for risc-v (and apparently a couple of others), but the s390x host is much slower apparently. I'm just going to disable the timeout for that test, and reenable it after the test.

the bytea failures are possibly an endian issue, but I'll know more once I have an s390x emulated locally.