Fix #75245: Don't set content of elements with only whitespaces

[eriklundin]

Bug #75245

Test code:
var_dump(simplexml_load_string('<test1><test2> </test2><test3></test3></test1>'));

Expected results (PHP 5.4.16)

object(SimpleXMLElement)#1 (2) {
  ["test2"]=>
  object(SimpleXMLElement)#2 (0) {
  }
  ["test3"]=>
  object(SimpleXMLElement)#3 (0) {
  }
}

Results from PHP 7.3.8+:

object(SimpleXMLElement)#1 (2) {
  ["test2"]=>
  object(SimpleXMLElement)#2 (1) {
    [0]=>
    string(4) "    "
  }
  ["test3"]=>
  object(SimpleXMLElement)#3 (0) {
  }
}

Results after patch:

object(SimpleXMLElement)#1 (2) {
  ["test2"]=>
  object(SimpleXMLElement)#2 (0) {
  }
  ["test3"]=>
  object(SimpleXMLElement)#3 (0) {
  }
}

[carusogabriel]

For reference, a running with all PHP versions: https://3v4l.org/02c8K

[saulery]

Good news !

Thanks a lot !

[nikic]

Which commit caused the original change in behavior?

[saulery]

Commit 406c215

Cf. https://github.com/php/php-src/blame/d891b5f458a253befaa56d325ae90518536fc2f0/ext/simplexml/simplexml.c#L1200

[nikic]

I've tried to understand how SXE dumping is supposed to work, but was left thoroughly confused. For example:

$sxe = simplexml_load_string('<root><test>x<test2>y</test2>z</test></root>');
var_dump($sxe);
// Gives:
object(SimpleXMLElement)#4 (1) {
  ["test"]=>
  string(2) "xz"
}

Is that supposed to happen, or is this a bug?

[eriklundin]

I've been looking at the spec for XML 1.0 and i can not find a reference to how tags after character data inside a tag should be handled. I'm sure it's not valid but how it should be handled by libxml/php i don't know. PHP seems to strip out the invalid data and leave the rest. This is possibly an acceptable solution.

However, if you would want the data with the invalid characters you would need to do something like this:

$sxe = simplexml_load_string('<root><test>x&lt;test2&gt;y&lt;/test2&gt;z</test></root>');
var_dump($sxe);

object(SimpleXMLElement)#1 (1) {
  ["test"]=>
  string(18) "x<test2>y</test2>z"
}

[krakjoe]

Can we get a status update here please ?

[eriklundin]

What's needed?

[krakjoe]

@eriklundin when I first read this I thought there were open questions, but I see now that you did answer ...

[krakjoe]

Merged as 6462c19

Thanks.

[nikic]

I've been looking at the spec for XML 1.0 and i can not find a reference to how tags after character data inside a tag should be handled. I'm sure it's not valid but how it should be handled by libxml/php i don't know. PHP seems to strip out the invalid data and leave the rest. This is possibly an acceptable solution.

Mixing character data and tags is valid XML ^^ I would expect this example to return something like

object(SimpleXMLElement)#%d (1) {
  ["test"]=>
  object(SimpleXMLElement)#%d (1) {
    [0]=>
    string(1) "x"
    ["test"] =>
    string(1) "y"
    [2]=>
    string(1) "z"
  }
}

or at least something with a similar information content. It seems pretty broken to me that SimpleXML just drops the inner tag entirely in this case.