Bug #68437 [Opn]: saveXML produces invalid xml

Edit report at https://bugs.php.net/bug.php?id=68437&edit=1 ID: 68437 Updated by: cmb@php.net Reported by: lphp-bug at thax dot hardliners dot org Summary: saveXML produces invalid xml Status: Open Type: Bug Package: DOM XML related PHP Version: master-Git-2014-11-17 (Git) Block user comment: N Private report: N New Comment: > XML (even the 1998 working drafts) defines valid chars as: > Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF] That was XML 1.0[1]. However, XML 1.1[2] defines: Char ::= [#x1-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF] [1] <http://www.w3.org/TR/REC-xml/#charsets> [2] <http://www.w3.org/TR/xml11/#charsets> Previous Comments: ------------------------------------------------------------------------ [2014-11-17 18:50:43] lphp-bug at thax dot hardliners dot org Description: ------------ Bug #54214 has been closed as Bogus, but gets the facts wrong: 1. XML (even the 1998 working drafts) defines valid chars as: Char ::= #x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF] This contradicts #54214 that Char ::= [#x1-#xD7FF] | ... Thus, 0x0b (or 0x06) are actually not in the valid Range. 2. libxml2 "problem" is not saving low-ascii chars, but loading them: loadXML(saveXML()) ALWAYS fails, when e.g. DOMText with low-ascii is present: DOMDocument::loadXML(): xmlParseCharRef: invalid xmlChar value 11 in Entity 3. saveXML() does not always warn. More specifically, it only warns when (*) $doc has no encoding AND (**) saveXML() is called without arguments. 4. When saveXML() does warn, it will remove the offending character (good), but not when it does not warn. Background: - saveXML() uses xmlDocDumpFormatMemory(), which uses document encoding, when given. Otherwise ascii encoding (0x00 ... 0x7f only) is assumed, and xmlEscapeEntities (libxml2/tree/xmlsave.c, line 208) is used as escape-function. This function warns and ignores the offending character - saveXML($el) uses xmlNodeDump(), which passes encoding==NULL to xmlNodeDumpOuput, which converts NULL to "UTF-8". Then xmlSaveCtxtInit() will not set the escape-function. When the escape-function is not set on the xmlSaveCtxt, xmlOutputBufferWriteEscape will use xmlEscapeContent (libxml2/tree/xmlIO.c, line 3536) instead. This function does not warn, and happily includes 0x0b in the output (where finally loadXML chokes on it). Libxml never says that it will warn for low-ascii chars. Thus php-developers can't depend on it. OTOH, it's unfeasible to force every php-programmer to pre-sanitize strings sent into DOMText, or post-check the saved xml string. Therefore php has to provide the necessary means to avoid invalid xmls -- i.e. at the very least warn, or even better: throw already in DOMText. Test script: --------------- $doc=new DOMDocument('1.0','utf-8'); // or: $doc=new DOMDocument('1.0'); (*) $el=$doc->createElement('root'); $doc->appendChild($el); $el->appendChild(new DOMText("\x0b asdf")); $str=$doc->saveXML($el); // or: $doc->saveXML() (**) var_dump($str); $doc->loadXML($str); Expected result: ---------------- loadXML shall load the saved xml, or throw an exception earlier. Actual result: -------------- loadXML only works when (*) and (**) are used, i.e. when saveXML outputs a warning. ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=68437&edit=1