Bug Report

[richard-guan-dev]

When I parse a string like "1", "12". JsonCPP just goes wrong.
parse function return true

[richard-guan-dev]

In face, if the string is begin with a number, the program will crash.

[cdunn2001]

Can you add a test-case to the unit-test suite?

[raandoom]

Hi. There is bug report about this: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=753560
This report contains patch and simple valid program for checking bug. Hope it helps.

[cdunn2001]

I don't follow the logic of the patch submitter. If the exception is thrown, then the type must be objectValue or arrayValue. If the parser is interpreting 9001 as one of those, then the bug seems to be in the parser. Can anyone prove and find that bug? (I don't have time.)

The patch would alter semantics. (Maybe null should never have been considered an array and an object, but it is.)

[raandoom]

Ok. I found something. This problem in Debian (maybe other distros) because it has old version of Value::asString() function in json_value.cpp. It looks like this:

std::string 
Value::asString() const
{
   switch ( type_ )
   {
   case nullValue:
      return "";
   case stringValue:
      return value_.string_ ? value_.string_ : "";
   case booleanValue:
      return value_.bool_ ? "true" : "false";
   case intValue:
   case uintValue:
   case realValue:
   case arrayValue:
   case objectValue:
      JSON_FAIL_MESSAGE( "Type is not convertible to string" );
   default:
      JSON_ASSERT_UNREACHABLE;
   }
   return ""; // unreachable
}

But current version use valueToString() function for types.

[cdunn2001]

So the current version is fine, right?

If you branch the Debian version and fix that, we'd be happy to accept that. But I doubt Debian would bother to pull the fix.

Close. (I love your username, btw.)