Hi John
Maybe it's a subtle platform-specific thing (I'm testing on macosx), but
the actual error is thrown because the scanner tries to read a directory
entry (the current working directory) -- and not a real file. Whenever
that value for the fExternalNoNamepaceSchemaLocation is a non-null
pointer, the scanner appears to construct a file url, concatenating the
current working directory with the file name (an empty string). So it
ends up trying to open and read a directory instead of a file (e.g it
tries to parse and read file:///user/jnapoli). I'll try on a different
platform to see if I still see a problem.
Thanks,
Jerry
John Snelson wrote:
Hi Jerry,
That stack trace appears to be from a different problem (possibly a file
not found?) - IGXMLScanner::scanTagStartNS doesn't appear in it anywhere.
I just tested XQilla using the simple-context-item.cpp sample which uses
the IGXMLScanner, and it worked fine for me.
John
Jerry Napoli wrote:
Hi John,
It looks like I was having the issue with IGXMLScanner. Here is the
stack trace:
#0 0x94d9a201 in __cxa_throw ()
#1 0x006c47ec in xercesc_2_8::XMLMacPosixFile::read at
MacPosixFile.cpp:147
#2 0x006bf244 in xercesc_2_8::XMLPlatformUtils::readFileBuffer at
MacOSPlatformUtils.cpp:268
#3 0x006091e8 in xercesc_2_8::BinFileInputStream::readBytes at
BinFileInputStream.cpp:102
#4 0x0077e7c0 in xercesc_2_8::XMLReader::refreshRawBuffer at
XMLReader.cpp:1699
#5 0x00781baa in xercesc_2_8::XMLReader::XMLReader at XMLReader.cpp:126
#6 0x00781cbf in xercesc_2_8::XMLReader::XMLReader at XMLReader.cpp:164
#7 0x006e807f in xercesc_2_8::ReaderMgr::createReader at
ReaderMgr.cpp:439
#8 0x006b5282 in xercesc_2_8::IGXMLScanner::scanReset at
IGXMLScanner2.cpp:1359
#9 0x006acaf1 in xercesc_2_8::IGXMLScanner::scanDocument at
IGXMLScanner.cpp:195
#10 0x005f4811 in xercesc_2_8::AbstractDOMParser::parse at
AbstractDOMParser.cpp:517
#11 0x006b3e05 in xercesc_2_8::IGXMLScanner::resolveSchemaGrammar at
IGXMLScanner2.cpp:1858
#12 0x006a7c89 in xercesc_2_8::IGXMLScanner::scanStartTagNS at
IGXMLScanner.cpp:2215
#13 0x006ac976 in xercesc_2_8::IGXMLScanner::scanContent at
IGXMLScanner.cpp:896
#14 0x006acb50 in xercesc_2_8::IGXMLScanner::scanDocument at
IGXMLScanner.cpp:214
#15 0x00d52ad9 in DocumentCacheImpl::parseDocument at
src/schema/DocumentCacheImpl.cpp:268
#16 0x00d52822 in DocumentCacheImpl::loadDocument at
src/schema/DocumentCacheImpl.cpp:253
#17 0x00d9b232 in FastXDMURIResolver::resolveDocument at
src/fastxdm/FastXDMConfiguration.cpp:91
#18 0x00bf3b70 in XQDynamicContextImpl::resolveDocument at
src/context/impl/XQDynamicContextImpl.cpp:337
#19 0x000023c1 in main at t_q2.cpp:17
It turns out the problem is in fact triggered by an empty string
passed to XMLScanner::setExternalNoNamespaceSchemaLocation. It falls
into this code in IGXMLScanner::scanTagStartNS (throwing at line 2215):
2209 if (isRoot && fDoSchema
2210 && (fExternalSchemaLocation ||
fExternalNoNamespaceSchemaLocation)) {
2211 2212 if (fExternalSchemaLocation)
2213 parseSchemaLocation(fExternalSchemaLocation);
2214 if (fExternalNoNamespaceSchemaLocation)
2215 resolveSchemaGrammar(fExternalNoNamespaceSchemaLocation,
XMLUni::fgZeroLenString);
2216 }
The code in my main() is right out of the XQilla examples:
// Parse a document, and set it as the context item
Sequence seq = context->resolveDocument(X("foo.xml"), 0);
Thanks,
Jerry
John Snelson wrote:
It seems to me that the SGXMLScanner is at fault here, since XQilla
works fine with the IGXMLScanner.
The problem seems to be caused because the IGXMLScanner uses a method
called processSchemaLocation() to parse the schema location string,
where as the SGXMLScanner uses XMLString::tokenizeString().
John
J Napoli wrote:
Hi,
While testing XQilla with Xerces-C 2.8.0 I was getting XML parse
exceptions trying to run the basic examples. Looking into the code
I noticed that XQilla by default passes an empty string (where val
!= NULL && XMLString::stringLen(val) == 0) to
XMLScanner::setExternalSchemaLocation and
setExternalNoNamespaceSchemaLocation. This seems to trick the
scanner into thinking there is an external schema to parse.
SGMLScanner::resolveSchemaGrammar then throws a parse exception
because it tries to parse a URL constructed by concatenating the
current working directory with the empty string.
Is this an XQilla or Xerces problem? Should Xerces ignore 0-length
strings here?
Modifying XMLScanner::setExternalSchemaLocation to ignore 0-length
strings seems to fix the problem (same issue with
setExternalNoNamespaceSchemaLocation):
----
inline void XMLScanner::setExternalSchemaLocation(const XMLCh* const
schemaLocation)
{
if (fExternalSchemaLocation) {
fMemoryManager->deallocate(fExternalSchemaLocation);
fExternalSchemaLocation = NULL;
}
// Only set if we're passed a non-trivial string
if (schemaLocation && XMLString::stringLen (schemaLocation))
fExternalSchemaLocation =
XMLString::replicate(schemaLocation, fMemoryManager);
}
Thanks,
Jerry
