On Sun, Jun 17, 2012 at 1:06 PM, Karoly Negyesi wrote:
PHP often throws fatal errors when it could throw an exception.
Take PHP Fatal error: Non-abstract method foo::bar() must contain
body this error for example. It would be possible to mark this class
as 'invalid' instead and when doing a new foo , throw an exception. Or
if you try really hard then only the method call should throw an
Or, I was extending ReflectionClass and got Fatal error:
ReflectionClass::getFileName(): Internal error: Failed to retrieve the
reflection object (same happens if you try serialzie / unserialzie on
ReflectionClass as a comment in the manual points out). I could argue
this is an exception and not a fatal.
Let me quote documentation:
Fatal run-time errors. These indicate errors that can not be recovered
from, such as a memory allocation problem. Execution of the script is
I do not see the above or , indeed , most of fatals I see this sort of
error. Sure, if a malloc call fails, the world has ended and it's time
to die. Otherwise, could we just get an exception so that it can be
Karoly "chx" Negyesi
PHP Internals - PHP Runtime Development Mailing List
To unsubscribe, visit: http://www.php.net/unsub.php
This kind of thing has come up before, without resolution. There are two
separate issues in your proposal:
1. Fatal errors that mean the script should stop executing, but don't
actually reflect instability in the engine.
2. Throwing exceptions from PHP core.
It seems to me that #2 is strongly opposed by a lot of people.
I think most people here sympathize with #1, but it hasn't been important
enough for anyone to actually do the work, and there hasn't been a proposal
about how it should be done. In general, it requires an audit of every
E_FATAL error thrown. Some can be changed to E_RECOVERABLE_ERROR. Some
however, really should kill the script, but don't need to skip error
handlers. These would need some new E_ constant, and a rule that when
thrown, an error handler can be called, but the script will die after the
error handler has finished. There's probably more to it than that, which is
why there would need to be a RFC. I'm guessing that if someone wrote an
RFC, and had a patch, it would be received well.