Bruno Haible via GNU coreutils General Discussion <[email protected]>
writes:

> The CI reports that the split/suffix-length test fails on macOS and on 
> OpenBSD.
>
> On macOS 26:
>
> FAIL: tests/split/suffix-length
> ===============================
> [...]
> + ulimit -v 6001
> ../tests/split/suffix-length.sh: line 81: ulimit: virtual memory: cannot 
> modify limit: Invalid argument
> + fail=1
>
> On OpenBSD 7.7:
>
> FAIL: tests/split/suffix-length
> ===============================
> [...]
> ../tests/split/suffix-length.sh: ulimit: -v: unknown option
> FAIL tests/split/suffix-length.sh (exit status: 1)

The 'ulimit -v' calls are missing a check to see if they are successful.

I pushed the attached patch after testing it fixed the issue on both
platforms.

Thanks,
Collin

>From fa1ad3fa07538fa3377498816ec9cf71039dd744 Mon Sep 17 00:00:00 2001
Message-ID: <fa1ad3fa07538fa3377498816ec9cf71039dd744.1781587418.git.collin.fu...@gmail.com>
From: Collin Funk <[email protected]>
Date: Mon, 15 Jun 2026 22:06:22 -0700
Subject: [PATCH] tests: split: avoid a false failure when ulimit fails

* tests/split/suffix-length.sh: Don't fail the test case if ulimit
fails.
Reported by Bruno Haible.
---
 tests/split/suffix-length.sh | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tests/split/suffix-length.sh b/tests/split/suffix-length.sh
index 0b2bf3894..5fca73ada 100755
--- a/tests/split/suffix-length.sh
+++ b/tests/split/suffix-length.sh
@@ -78,7 +78,10 @@ test -f xaa && fail=1
 rm -f x*
 vm=$(get_min_ulimit_v_ split -a 1 /dev/null)
 if test -n "$vm"; then
-  (ulimit -v $(($vm+6000)) && returns_ 1 split -a 66542562175252 in) || fail=1
+  (ulimit -v $(($vm+6000)) &&
+   touch ulimit-worked &&
+   returns_ 1 split -a 66542562175252 in) ||
+    { test -f ulimit-worked && fail=1; }
   test -f xaa && fail=1
 fi
 
-- 
2.54.0

Reply via email to