Introduce FX_Number class as a replacement for FX_atonum().

The issue with FX_atonum() is that it doesn't return any information
about whether it range-checked its integer values as a signed or
unsigned type, even though it knows this as part of its processing.

Rather than adding another out parameter to that function, create
a class to hold all this information together.

This is the first place things went astray while diagnosing
bug 882959, in that a large positive value was cast to float as a
negative value. Unfortunately, this doesn't affect the related bug,
but is a step in the right direction.

Change-Id: I0977ec8fccf85e2632a962507bdd30a1cbe6d33c
Reviewed-on: https://pdfium-review.googlesource.com/42353
Reviewed-by: Lei Zhang <thestig@chromium.org>
Commit-Queue: Tom Sepez <tsepez@chromium.org>

1 files changed, 4 insertions, 7 deletions

diff --git a/core/fpdfapi/parser/cpdf_number.h b/core/fpdfapi/parser/cpdf_number.h
index b5faf82bb2..724ae9f1f1 100644
--- a/core/fpdfapi/parser/cpdf_number.h
+++ b/core/fpdfapi/parser/cpdf_number.h
@@ -10,6 +10,7 @@
 #include <memory>
 
 #include "core/fpdfapi/parser/cpdf_object.h"
+#include "core/fxcrt/fx_number.h"
 #include "core/fxcrt/fx_string.h"
 #include "core/fxcrt/fx_system.h"
 
@@ -34,14 +35,10 @@ class CPDF_Number final : public CPDF_Object {
   bool WriteTo(IFX_ArchiveStream* archive,
                const CPDF_Encryptor* encryptor) const override;
 
-  bool IsInteger() const { return m_bInteger; }
+  bool IsInteger() const { return m_Number.IsInteger(); }
 
- protected:
-  bool m_bInteger;
-  union {
-    int m_Integer;
-    float m_Float;
-  };
+ private:
+  FX_Number m_Number;
 };
 
 inline CPDF_Number* ToNumber(CPDF_Object* obj) {