Ref #8202

http://www.php.net/manual/en/function.exit.php

Exit statuses should be in the range 0 to 254, the exit status 255 is reserved by PHP and shall not be used. The status 0 is used to terminate the program successfully.

So in php terms we would need to mask with 254 which would also not be consistent with linux.

@Tobion It's been a while I found a practical use for bit-operations, but if I remember correctly masking with 254 would mean, that you only can set even numbers, because the lowest bit is missing (011 & 010 = 010)

It seems that a simple php script:

  php -r 'exit(300);' ; echo $?

actually outputs 44 as well (strace shows exit_group(300), so php does not cap either).

Maybe it would even be better to not handle these situations after all, and remove the cap all together without replacing it with a mask?

@jaytaph If PHP is doing the right thing for us, let's remove the logic on our side.

@fabpot I think that is a good idea.. however, I do not know what will happen on windows machines. Maybe somebody should verify if this behaviour is the same?

@jaytaph on windows it's not masked. The code below returns 300. So I guess the masking is done by the linux console and not by php because in mingw your example also outputs 44.

php -r "exit(300);"
echo %ERRORLEVEL%

@Tobion The cap is actually done by the linux exit_group system call: http://lxr.free-electrons.com/source/kernel/exit.c#L842 Here is the exit code masked with 0xFF (and shifted, because the lower bits have different meanings, like core dumped,stopped, continued etc).

Same exit result (44) works on Darwin too.

It seems that under windows, exit codes are 32-bit signed integers (http://en.wikipedia.org/wiki/Exit_status#Windows), so here no filtering should take place anyway (an exit(300) seems valid).

Maybe we should thus get rid of the cap, and not even replace it with a mask? As every supported OS seems to handle this fine, and PHP doesn't really care either way (exit codes are passed directly to the OS), even the 255 reserved code seems to cause no issues (a quick scan through the php core doesn't even seem to suggest 255 should be considered reserved, maybe this is a remnant from past versions).

Not capping has a severe issue: exit(256) will then become a 0 exit code on Linux, which is a success one. This is bad (especially as we can use the exception code as exit code)

Closing it for the reasons given by @stof in the previous comment and @Tobion in #17591 (comment).