From d6baa3cab0271ec7c2a92f7d3f51dc7242330292 Mon Sep 17 00:00:00 2001
From: Spike Curtis <spike@coder.com>
Date: Fri, 26 Jan 2024 19:22:35 +0400
Subject: [PATCH] fix: stop running tests that exec sh scripts in parallel
 (#11834)

Ok, so my last attempt at a fix here failed

https://github.com/coder/coder/actions/runs/7666229961/job/20893608286

I have a new theory: it's not the `terraform` binary that's busy, it's actually `fake_cancel.sh` and it gets marked busy when we `exec` it from the script we write.

Use of `exec` also replaces the executing code in place, rather than starting a new process/shell, so that's why the error we get says `terraform` is busy.
---
 provisioner/terraform/provision_test.go | 25 ++++++++++---------------
 1 file changed, 10 insertions(+), 15 deletions(-)

diff --git a/provisioner/terraform/provision_test.go b/provisioner/terraform/provision_test.go
index 85868fe611..1395877b18 100644
--- a/provisioner/terraform/provision_test.go
+++ b/provisioner/terraform/provision_test.go
@@ -14,7 +14,6 @@ import (
 	"runtime"
 	"sort"
 	"strings"
-	"syscall"
 	"testing"
 	"time"
 
@@ -126,8 +125,10 @@ func sendApply(sess proto.DRPCProvisioner_SessionClient, transition proto.Worksp
 	}}})
 }
 
+// below we exec fake_cancel.sh, which causes the kernel to execute it, and if more than
+// one process tries to do this simultaneously, it can cause "text file busy"
+// nolint: paralleltest
 func TestProvision_Cancel(t *testing.T) {
-	t.Parallel()
 	if runtime.GOOS == "windows" {
 		t.Skip("This test uses interrupts and is not supported on Windows")
 	}
@@ -158,24 +159,16 @@ func TestProvision_Cancel(t *testing.T) {
 	}
 	for _, tt := range tests {
 		tt := tt
+		// below we exec fake_cancel.sh, which causes the kernel to execute it, and if more than
+		// one process tries to do this, it can cause "text file busy"
+		// nolint: paralleltest
 		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
 			dir := t.TempDir()
 			binPath := filepath.Join(dir, "terraform")
 
 			// Example: exec /path/to/terrafork_fake_cancel.sh 1.2.1 apply "$@"
 			content := fmt.Sprintf("#!/bin/sh\nexec %q %s %s \"$@\"\n", fakeBin, terraform.TerraformVersion.String(), tt.mode)
-
-			// golang's standard OS library can sometimes leave the file descriptor open even after
-			// "Closing" the file (which can then lead to a "text file busy" error, so we bypass this
-			// and use syscall directly).
-			fd, err := syscall.Open(binPath, syscall.O_WRONLY|syscall.O_CREAT, 0o755)
-			require.NoError(t, err)
-			n, err := syscall.Write(fd, []byte(content))
-			require.NoError(t, err)
-			require.Equal(t, len(content), n)
-			err = syscall.Close(fd)
+			err := os.WriteFile(binPath, []byte(content), 0o755) //#nosec
 			require.NoError(t, err)
 			t.Logf("wrote fake terraform script to %s", binPath)
 
@@ -228,8 +221,10 @@ func TestProvision_Cancel(t *testing.T) {
 	}
 }
 
+// below we exec fake_cancel_hang.sh, which causes the kernel to execute it, and if more than
+// one process tries to do this, it can cause "text file busy"
+// nolint: paralleltest
 func TestProvision_CancelTimeout(t *testing.T) {
-	t.Parallel()
 	if runtime.GOOS == "windows" {
 		t.Skip("This test uses interrupts and is not supported on Windows")
 	}