From 5bf4908057d611a127020fca7acf35c606cade16 Mon Sep 17 00:00:00 2001
From: Eric Wong <normalperson@yhbt.net>
Date: Sun, 29 Mar 2009 18:59:48 -0700
Subject: test_exec: fix race conditions

We need to ensure the QUIT signal to the old processes
are processed when fixing the config.

Additionally, the log rotation checker was not reliable
because the master log emitted a similar message to
the workers and we were not distinguishing between
them.  Check for all 5 logs (1 master + 4 workers)
to be rotated.
---
 test/exec/test_exec.rb | 55 +++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 41 insertions(+), 14 deletions(-)

(limited to 'test/exec')

diff --git a/test/exec/test_exec.rb b/test/exec/test_exec.rb
index 271ae20..fd33db1 100644
--- a/test/exec/test_exec.rb
+++ b/test/exec/test_exec.rb
@@ -128,7 +128,9 @@ end
     Process.kill(:USR2, File.read(pid_file).to_i)
     wait_for_file(old_file)
     wait_for_file(pid_file)
-    Process.kill(:QUIT, File.read(old_file).to_i)
+    old_pid = File.read(old_file).to_i
+    Process.kill(:QUIT, old_pid)
+    wait_for_death(old_pid)
 
     ucfg.syswrite("timeout %(#{pid_file})\n") # introduce a bug
     current_pid = File.read(pid_file).to_i
@@ -158,7 +160,8 @@ end
     ucfg.syswrite("listen %(#@addr:#@port)\n")
     ucfg.syswrite("listen %(#@addr:#{port2})\n")
     ucfg.syswrite("pid %(#{pid_file})\n")
-    Process.kill(:USR2, current_pid)
+    assert_nothing_raised { Process.kill(:USR2, current_pid) }
+
     wait_for_file(old_file)
     wait_for_file(pid_file)
     new_pid = File.read(pid_file).to_i
@@ -195,7 +198,9 @@ end
     Process.kill(:USR2, File.read(pid_file).to_i)
     wait_for_file(old_file)
     wait_for_file(pid_file)
-    Process.kill(:QUIT, File.read(old_file).to_i)
+    old_pid = File.read(old_file).to_i
+    Process.kill(:QUIT, old_pid)
+    wait_for_death(old_pid)
 
     File.unlink("config.ru") # break reloading
     current_pid = File.read(pid_file).to_i
@@ -220,7 +225,7 @@ end
 
     # fix the bug
     File.open("config.ru", "wb") { |fp| fp.syswrite(HI) }
-    Process.kill(:USR2, current_pid)
+    assert_nothing_raised { Process.kill(:USR2, current_pid) }
     wait_for_file(old_file)
     wait_for_file(pid_file)
     new_pid = File.read(pid_file).to_i
@@ -293,20 +298,21 @@ end
     # USR1 should've been passed to all workers
     tries = DEFAULT_TRIES
     log = File.readlines(rotate.path)
-    while (tries -= 1) > 0 && log.grep(/rotating logs\.\.\./).size < 4
+    while (tries -= 1) > 0 &&
+          log.grep(/rotating logs\.\.\./).size < 5
       sleep DEFAULT_RES
       log = File.readlines(rotate.path)
     end
-    assert_equal 4, log.grep(/worker=\d+ rotating logs\.\.\./).size
+    assert_equal 5, log.grep(/rotating logs\.\.\./).size
     assert_equal 0, log.grep(/done rotating logs/).size
 
     tries = DEFAULT_TRIES
     log = File.readlines(COMMON_TMP.path)
-    while (tries -= 1) > 0 && log.grep(/done rotating logs/).size < 4
+    while (tries -= 1) > 0 && log.grep(/done rotating logs/).size < 5
       sleep DEFAULT_RES
       log = File.readlines(COMMON_TMP.path)
     end
-    assert_equal 4, log.grep(/worker=\d+ done rotating logs/).size
+    assert_equal 5, log.grep(/done rotating logs/).size
     assert_equal 0, log.grep(/rotating logs\.\.\./).size
     assert_nothing_raised { Process.kill(:QUIT, pid) }
     status = nil
@@ -529,15 +535,18 @@ end
       wait_for_file("#{pid_file}.oldbin")
       wait_for_file(pid_file)
 
+      old_pid = File.read("#{pid_file}.oldbin").to_i
+      new_pid = File.read(pid_file).to_i
+
       # kill old master process
-      assert_not_equal pid, File.read(pid_file).to_i
-      assert_equal pid, File.read("#{pid_file}.oldbin").to_i
-      assert_nothing_raised { Process.kill(:QUIT, pid) }
-      assert_not_equal pid, File.read(pid_file).to_i
+      assert_not_equal pid, new_pid
+      assert_equal pid, old_pid
+      assert_nothing_raised { Process.kill(:QUIT, old_pid) }
       assert_nothing_raised { retry_hit(["http://#{@addr}:#{@port}/"]) }
-      wait_for_file(pid_file)
+      wait_for_death(old_pid)
+      assert_equal new_pid, File.read(pid_file).to_i
       assert_nothing_raised { retry_hit(["http://#{@addr}:#{@port}/"]) }
-      assert_nothing_raised { Process.kill(:QUIT, File.read(pid_file).to_i) }
+      assert_nothing_raised { Process.kill(:QUIT, new_pid) }
     end
 
     def reexec_basic_test(pid, pid_file)
@@ -580,4 +589,22 @@ end
       end
     end
 
+    # can't waitpid on detached processes
+    def wait_for_death(pid)
+      tries = DEFAULT_TRIES
+      while (tries -= 1) > 0
+        begin
+          Process.kill(0, pid)
+          begin
+            Process.waitpid(pid, Process::WNOHANG)
+          rescue Errno::ECHILD
+          end
+          sleep(DEFAULT_RES)
+        rescue Errno::ESRCH
+          return
+        end
+      end
+      raise "PID:#{pid} never died!"
+    end
+
 end if do_test
-- 
cgit v1.2.3-24-ge0c7