Bug #74569 [Com]: Return in a finally clause silently ignores an exception thrown in a try clause

From: Date: Fri, 12 May 2017 18:24:18 +0000
Subject: Bug #74569 [Com]: Return in a finally clause silently ignores an exception thrown in a try clause
References: 1  Groups: php.bugs 
Request: Send a blank email to php-bugs+get-209087@lists.php.net to get a copy of this message
Edit report at https://bugs.php.net/bug.php?id=74569&edit=1 ID: 74569 Comment by: trianman at gmail dot com Reported by: trianman at gmail dot com Summary: Return in a finally clause silently ignores an exception thrown in a try clause Status: Open Type: Bug Package: Scripting Engine problem Operating System: Any PHP Version: 7.1.4 Block user comment: N Private report: N New Comment: 2 danack@php.net Sorry, I don't want to be mean. My English skills are not very good and it is sometimes hard to express my thoughts in a clear way. I've just realized that the finally statement is some-kind of antipattern for me. Because finally is a shortcut for catching and re-throwing an \Exception class eg. <?php try { throw new CustomException(); } catch (CustomException $ex) { handleException(); } catch (\Exception $ex) { doThingsInFinally(); throw $ex; } doThingsInFinally(); ?> Or with finally: <?php try { throw new CustomException(); } catch (CustomException $ex) { handleException(); } finally { doThingsInFinally(); } ?> Of cause if we have a return statement instead of a doTihingsInFinally() method, we will lose an exception. But we doing it in a clear way. On the other hand finally statement hides out from developer's eyes what actually is going. So for my mind a little copy-paste is a lesser evil than an unclear behavior. For all other guys the E_NOTICE will be enough. (: Previous Comments: ------------------------------------------------------------------------ [2017-05-11 14:28:34] olavisau at gmail dot com BC is broken with any solution. The fact that javascript behaves in the same way worries me. It's probably very hard to implement. I agree with the E_NOTICE. ------------------------------------------------------------------------ [2017-05-11 14:21:03] trianman at gmail dot com > btw The behaviour in PHP is the same as that which happens in Java > This behaviour is also that used in Javascript. Sorry to hear that, but AFAIK nor Java nor JavaScript (ECMAScript) has no conventional (standartized) way to produce non critical warnings at the run time. So PHP is better in this way and why can't we employ it? > "Don't put returns in finally" I believe that the good tool must help its users not to "shoot themselves" as much as possible. So the user becomes free from a lot of "Don't do something somewhere". If you don't agree you are free to use an assembler to do anything you want and remember a lot of such restrictions. I can imagine another more expectable an clear way to handle this case. But it will break a BC so can be implemented only in a new major release of PHP. Triggering E_NOTICE is the most convenient way for now. ------------------------------------------------------------------------ [2017-05-11 13:42:04] danack@php.net This behaviour is also that used in Javascript. function foo() { try { throw 42; } finally { return "edge cases are hard"; } }; alert(foo()); ------------------------------------------------------------------------ [2017-05-11 13:30:49] danack@php.net > but it MUST NOT be silently dropped out. That's a conclusion without rigorous justification. btw The behaviour in PHP is the same as that which happens in Java: https://web.archive.org/web/20070922061412/http://weblogs.java.net/blog/staufferjames/archive/2007/06/_dont_return_in.html It'd be a lot easier to say "Don't put returns in finally blocks, yo". > The interpreter should trigger an E_NOTICE or something similar here. This would almost certainly be better done with code analysis tools, rather than giving warnings on valid (though highly surprising) code. ------------------------------------------------------------------------ [2017-05-11 07:41:34] olavisau at gmail dot com The correct way of writing this would be to catch the exception and throw it again, however this also fails. If it is not thrown again, then the behavior is correct. <?php class MyException extends Exception { public function __construct($message = "", $code = 0, Throwable $previous = null) { parent::__construct($message, $code, $previous); print "2. Exception constructed\n"; } } function throw_exception() { try { throw new MyException("5. Exception message\n"); } catch(Exception $exception) { print "3. Exception caught and thrown again\n"; throw $exception; } finally { return "4. Return statement\n"; } } try { print "1. Start\n"; print throw_exception(); } catch (Exception $e) { print $e->getMessage(); } The flow should be: exception is thrown, caught by the inner block, thrown again, caught by the outer block and then the finally block should execute, returning the flow to print throw_exception(); ------------------------------------------------------------------------ The remainder of the comments for this report are too long. To view the rest of the comments, please view the bug report online at https://bugs.php.net/bug.php?id=74569 -- Edit this bug report at https://bugs.php.net/bug.php?id=74569&edit=1

« previous php.bugs (#209087) next »