Reliance on pcntl_fork about boris HOT 7 CLOSED

Thanks for the report. This is documented in the README and is fundamental to the architecture, but I'd love to see any suggestions for a non-forking solution that is able to handle fatal errors. Boris is different to other PHP REPLs, in that it was built specifically to deal with fatal errors, which basically means each statement must be executed in a new process, in case it dies. If you don't care about the REPL quitting on fatal errors, you may be interested in https://github.com/ieure/php_repl.

I'll leave this ticket open in case somebody can come up with an approach that doesn't rely on the ability to fork.

from boris.

This isn't a way around it but a way to get it running if you have access to php.ini. I used this guide to get pcntl running for MAMP on OS X.

http://stackoverflow.com/questions/5230255/how-to-enable-process-control-extension-pcntl-in-php-mamp

from boris.

Thanks for this. The idea of the forking is fundamental to how boris deals with fatal errors (actually, you may want to try since I reverted that last commit just now, if you found boris exited on error... I just fixed an issue).

I'm still keen to find a solution that works on Windows :)

from boris.

Not a problem. Windows, though. Oooh boy.

from boris.

I don't think the reliance on the process control functions or the posix functions can be avoided at all at this stage... Boris simply could not exist without them. I may consider writing an installer script that tries to install what it needs from PECL, however. It may also be possible to package Boris up into a Debian package with dependencies formally specified, so they can be installed.

from boris.

I ran into the same problem. The following article is very helpful with enabling pcntl on a default Mac OS X php installation:

http://moffe42.blogspot.com.au/2012/06/installing-pcntl-for-php-on-osx-lion.html

from boris.

I'll leave this ticket open in case somebody can come up with an approach that doesn't rely on the ability to fork.

One alternative to pcntl_fork is to "recreate" the current PHP environment by starting a new PHP in which you load all the same code and update all global state to match the current state. PHPUnit does this when you configure runTestInSeparateProcess, to avoid a dependency on fork.
See sebastianbergmann/phpunit#909
and /src/Framework/TestCase.php:783

This is somewhat fragile, and more likely to work for something like PHPUnit which is running from PHP source files, than for Boris which is running from user input. Further, if there is any code which has side-effects, it may get re-run on each pseudo-fork.

from boris.