Add $result_code parameter to shell_exec

[l-pt]

RFC: https://wiki.php.net/rfc/shell_exec_result_code_param

[ricardoboss]

Won't this evaluate to false if ret_code === 0?

[jbrooksuk]

If so, is exec not also affected by this?

php-src/ext/standard/exec.c

Lines 246 to 248 in 41d2785

	if (ret_code) {
	ZEND_TRY_ASSIGN_REF_LONG(ret_code, ret);
	}

[guilliamxavier]

TL;DR: this code is fine.
[First, that's C not PHP, so there's no "===" operator; second,] ret_code is a zval * i.e. a pointer to zval, so if (ret_code) means if (ret_code != NULL) i.e. (here) the PHP function was called with a 2nd argument (a variable into which to copy the value of ret_pclose).

[nikic]

This needs some tests.

[drupol]

Sorry to reply here, I'm not subscribed to the mailing list yet, will fix that during the weekend.

I added a 👎 because I think that adding referenced parameters adds confusion everywhere they are used.

The feature is definitely useful, but I think it should be achieved in a different way, in a more "pure" (in terms of function).

There are many options for that:

• Returning a structured array - this would be a definitive major BC break and an ugly solution.
• Returning a Stringable structural object - This would be a better solution and a minor BC break
• Other ideas?

[divinity76]

isn't this 127 on unix-like OS's (Linux, MacOS, BSD),
but just 1 on Windows ? then again i wouldn't expect to find /bin/true on Windows

it is possible to make /bin/true executable on windows tho:

C:\>mkdir bin

C:\>cd bin

C:\bin>copy C:\Windows\System32\PING.EXE .
        1 file(s) copied.

C:\bin>move PING.EXE true.exe
        1 file(s) moved.

C:\bin>php -r "var_dump(is_executable('/bin/true'));"
bool(false)

C:\bin>move true.exe true
        1 file(s) moved.

C:\bin>php -r "var_dump(is_executable('/bin/true'));"
bool(true)

[cmb69]

What's the status here? The respective RFC is still listed as being under discussion, although apparently there was no more discussion for quite a while.

Are you planning to continue the RFC process for this feature, @l-pt?

[cmb69]

Actually, I don't see the point getting the result code with shell_exec(). It is already possible to use exec() instead, and since shell_exec() is equivalent to the backtick operator, how could this parameter be mapped to the operator?

[guilliamxavier]

@cmb69: A point argued in https://bugs.php.net/bug.php?id=81493 was:

to use exec() for the same task as shell_exec(), you'll have to (ab)use implode() like

exec($cmd,$output,$ret);
$output=implode("\n", $output);

and to use system() for the same job, you'd have to (ab)use ob_* like

ob_start();
system($cmd,$ret);
$output=ob_get_clean();

same for passthru()

and https://externals.io/message/116439 added:

makes sense for consistency

and I wonder if "Execute an external program" is really always identical to "Execute command via shell" (e.g. for shell builtins)?

As for the backtick operator, it is explicitly said to remain unaffected.

[cmb69]

and I wonder if "Execute an external program" is really always identical to "Execute command via shell" (e.g. for shell builtins)?

exec(), system(), passthru() and shell_exec() are wrappers over popen(3); as such, they should be interchangeable (except for the details you've mentioned, and additionally the use of different text/binary modes on Windows).