watchdog() mishandling $in_error_state

Code correctness.

The following is what you have now. Simplified. If you cannot see the BIG error, then I cannot help. That sample will run. I'm not too sure when it will stop however. The string "I'm running the loop!", according to the comment in the real function, will be written once. Will it?

function run_only_once() {
   static $avoid_recursivity = FALSE;

   if(!$avoid_recursivity) {
    $avoid_recursivity = TRUE;

    echo "I'm running the loop!";
    for($i = 1; $i < 10; ++$i) {
      run_only_once();
    }
  }
  $avoid_recursivity = FALSE;
}

Code correctness in the entire system is what makes the system work better. And the idea of not fixing something because no one has a problem with it, even though it is wrong, does not make sense, does it?

Thank you
Alexis Wilke

Dave,

I'm sorry you cannot see it.

I'll try again...

Your example does prevent the loop only on the 1st error, not the following ones. Why can't you see it?!

function mywatchdog() {
  static $avoid_recursivity = FALSE;

  if(!$avoid_recursivity) {
    $avoid_recursivity = TRUE;

    echo "EXECUTING WATCHDOG";

    mywatchdog_watchdog1();
    mywatchdog_watchdog2();
  }
  $avoid_recursivity = FALSE;
}

function mywatchdog_watchdog1() {
  // Cause an error, run watchdog
  mywatchdog();
  echo "watchdog1\n";
}

function mywatchdog_watchdog2() {
  // Cause an error, run watchdog
  mywatchdog();
  echo "watchdog2\n";
}

// Cause an error, run watchdog
mywatchdog();

And this time the result is something a bit more complex...

EXECUTING WATCHDOG
watchdog1
EXECUTING WATCHDOG
watchdog1
... [continuing until PHP crash]

Let me know if you find a different result.

And of course, feel free to move the line were we put the variable back to FALSE inside the if() to see the difference. Now this may already have caused mayhem to a few people...

kbahey, I do not know what programming theory you are talking about, but you may want to test the code properly: TWICE instead of ONCE. And my sample has no threading issue. Actually, that code, even fixed, would NOT work in a multi-threaded environment. For that you'd need a lock. But that's fine, PHP is not multi-threaded anyway.

Ha! Yes... not even in D6... 8-)

I'm glad you understand in the end.

Thank you.
Alexis

Hi Khalid,

I noticed that this was still open and now the computer tells me that the patch is not passing some tests.

Am I supposed to do something about it?!

Thank you.
Alexis

Hey!

That's it 8-) A 344Kb patch... 8-) (that DB the new generation node)

You see, the author accepted and it is one line that was at the wrong place, he even ran a test to see the problem and it does not get fixed and back ported immediately. It would take under 1h altogether and one less bug in the whole system. Then you avoid the 344Kb patches that fixes all these bugs at once... Okay, j/k, I just won't have the time to read 302 messages. But I can see that the problem is quite different. I just think that the process is very slow for such simple patches to be applied.

Thank you for posting a new patch and letting me know that the system has a problem because all my patches are generated failures now... This was not the only one!

Best,
Alexis