cleaning up common OSX code

[wxWidgets.git] / src / common / string.cpp

diff --git a/src/common/string.cpp b/src/common/string.cpp
index fbfe6cb6b5dfe16e641c434bcf5aa3855cabf9f1..6b6582462259fcedddf1c7f99c4f158dc9339b26 100644 (file)
--- a/src/common/string.cpp
+++ b/src/common/string.cpp
@@ -35,10 +35,6 @@
 #include <string.h>
 #include <stdlib.h>
 
 #include <string.h>
 #include <stdlib.h>
 

-#ifdef __SALFORDC__
-    #include <clib.h>
-#endif
-

 #include "wx/hashmap.h"
 
 // string handling functions used by wxString:
 #include "wx/hashmap.h"
 
 // string handling functions used by wxString:


@@ -72,11 +68,10 @@ const size_t wxString::npos = (size_t) -1;
 
 wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str)
 {
 
 wxSTD ostream& operator<<(wxSTD ostream& os, const wxCStrData& str)
 {

-// FIXME-UTF8: always, not only if wxUSE_UNICODE
-#if wxUSE_UNICODE && !defined(__BORLANDC__)
-    return os << (const wchar_t*)str.AsWCharBuf();
+#if wxUSE_UNICODE && !wxUSE_UNICODE_UTF8
+    return os << (const char *)str.AsCharBuf();

 #else
 #else

-    return os << (const char*)str.AsCharBuf();
+    return os << str.AsInternal();

 #endif
 }
 
 #endif
 }
 


@@ -97,6 +92,25 @@ wxSTD ostream& operator<<(wxSTD ostream& os, const wxWCharBuffer& str)
 }
 #endif
 
 }
 #endif
 

+#if wxUSE_UNICODE && defined(HAVE_WOSTREAM)
+
+wxSTD wostream& operator<<(wxSTD wostream& wos, const wxString& str)
+{
+    return wos << str.wc_str();
+}
+
+wxSTD wostream& operator<<(wxSTD wostream& wos, const wxCStrData& str)
+{
+    return wos << str.AsWChar();
+}
+
+wxSTD wostream& operator<<(wxSTD wostream& wos, const wxWCharBuffer& str)
+{
+    return wos << str.data();
+}
+
+#endif  // wxUSE_UNICODE && defined(HAVE_WOSTREAM)
+

 #endif // wxUSE_STD_IOSTREAM
 
 // ===========================================================================
 #endif // wxUSE_STD_IOSTREAM
 
 // ===========================================================================


@@ -1329,9 +1343,9 @@ wxString& wxString::MakeLower()
 // ---------------------------------------------------------------------------
 
 // some compilers (VC++ 6.0 not to name them) return true for a call to
 // ---------------------------------------------------------------------------
 
 // some compilers (VC++ 6.0 not to name them) return true for a call to

-// isspace('ê') in the C locale which seems to be broken to me, but we have to
-// live with this by checking that the character is a 7 bit one - even if this
-// may fail to detect some spaces (I don't know if Unicode doesn't have
+// isspace('\xEA') in the C locale which seems to be broken to me, but we have
+// to live with this by checking that the character is a 7 bit one - even if
+// this may fail to detect some spaces (I don't know if Unicode doesn't have

 // space-like symbols somewhere except in the first 128 chars), it is arguably
 // still better than trimming away accented letters
 inline int wxSafeIsspace(wxChar ch) { return (ch < 127) && wxIsspace(ch); }
 // space-like symbols somewhere except in the first 128 chars), it is arguably
 // still better than trimming away accented letters
 inline int wxSafeIsspace(wxChar ch) { return (ch < 127) && wxIsspace(ch); }


@@ -1426,61 +1440,62 @@ int wxString::Find(wxUniChar ch, bool bFromEnd) const
     #define DO_IF_NOT_WINCE(x)
 #endif
 
     #define DO_IF_NOT_WINCE(x)
 #endif
 

-#define WX_STRING_TO_INT_TYPE(val, base, func)                              \
-    wxCHECK_MSG( val, false, _T("NULL output pointer") );                   \
+#define WX_STRING_TO_INT_TYPE(out, base, func, T)                           \
+    wxCHECK_MSG( out, false, _T("NULL output pointer") );                   \

     wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") );  \
                                                                             \
     DO_IF_NOT_WINCE( errno = 0; )                                           \
                                                                             \
     const wxStringCharType *start = wx_str();                               \
     wxStringCharType *end;                                                  \
     wxASSERT_MSG( !base || (base > 1 && base <= 36), _T("invalid base") );  \
                                                                             \
     DO_IF_NOT_WINCE( errno = 0; )                                           \
                                                                             \
     const wxStringCharType *start = wx_str();                               \
     wxStringCharType *end;                                                  \

-    *val = func(start, &end, base);                                         \
+    T val = func(start, &end, base);                                        \

                                                                             \
     /* return true only if scan was stopped by the terminating NUL and */   \
     /* if the string was not empty to start with and no under/overflow */   \
     /* occurred: */                                                         \
                                                                             \
     /* return true only if scan was stopped by the terminating NUL and */   \
     /* if the string was not empty to start with and no under/overflow */   \
     /* occurred: */                                                         \

-    return !*end && (end != start)                                          \
-        DO_IF_NOT_WINCE( && (errno != ERANGE) )
+    if ( *end || end == start DO_IF_NOT_WINCE(|| errno == ERANGE) )         \
+        return false;                                                       \
+    *out = val;                                                             \
+    return true

 
 

-bool wxString::ToLong(long *val, int base) const
+bool wxString::ToLong(long *pVal, int base) const

 {
 {

-    WX_STRING_TO_INT_TYPE(val, base, wxStrtol);
+    WX_STRING_TO_INT_TYPE(pVal, base, wxStrtol, long);

 }
 
 }
 

-bool wxString::ToULong(unsigned long *val, int base) const
+bool wxString::ToULong(unsigned long *pVal, int base) const

 {
 {

-    WX_STRING_TO_INT_TYPE(val, base, wxStrtoul);
+    WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoul, unsigned long);

 }
 
 }
 

-bool wxString::ToLongLong(wxLongLong_t *val, int base) const
+bool wxString::ToLongLong(wxLongLong_t *pVal, int base) const

 {
 {

-    WX_STRING_TO_INT_TYPE(val, base, wxStrtoll);
+    WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoll, wxLongLong_t);

 }
 
 }
 

-bool wxString::ToULongLong(wxULongLong_t *val, int base) const
+bool wxString::ToULongLong(wxULongLong_t *pVal, int base) const

 {
 {

-    WX_STRING_TO_INT_TYPE(val, base, wxStrtoull);
+    WX_STRING_TO_INT_TYPE(pVal, base, wxStrtoull, wxULongLong_t);

 }
 
 }
 

-bool wxString::ToDouble(double *val) const
+bool wxString::ToDouble(double *pVal) const

 {
 {

-    wxCHECK_MSG( val, false, _T("NULL pointer in wxString::ToDouble") );
+    wxCHECK_MSG( pVal, false, _T("NULL output pointer") );

 
 

-#ifndef __WXWINCE__
-    errno = 0;
-#endif
+    DO_IF_NOT_WINCE( errno = 0; )

 
     const wxChar *start = c_str();
     wxChar *end;
 
     const wxChar *start = c_str();
     wxChar *end;

-    *val = wxStrtod(start, &end);
+    double val = wxStrtod(start, &end);

 
     // return true only if scan was stopped by the terminating NUL and if the
     // string was not empty to start with and no under/overflow occurred
 
     // return true only if scan was stopped by the terminating NUL and if the
     // string was not empty to start with and no under/overflow occurred

-    return !*end && (end != start)
-#ifndef __WXWINCE__
-        && (errno != ERANGE)
-#endif
-    ;
+    if ( *end || end == start DO_IF_NOT_WINCE(|| errno == ERANGE) )
+        return false;
+
+    *pVal = val;
+
+    return true;

 }
 
 // ---------------------------------------------------------------------------
 }
 
 // ---------------------------------------------------------------------------


@@ -1572,6 +1587,60 @@ int wxString::DoPrintfUtf8(const char *format, ...)
 }
 #endif // wxUSE_UNICODE_UTF8
 
 }
 #endif // wxUSE_UNICODE_UTF8
 

+/*
+    Uses wxVsnprintf and places the result into the this string.
+
+    In ANSI build, wxVsnprintf is effectively vsnprintf but in Unicode build
+    it is vswprintf.  Due to a discrepancy between vsnprintf and vswprintf in
+    the ISO C99 (and thus SUSv3) standard the return value for the case of
+    an undersized buffer is inconsistent.  For conforming vsnprintf
+    implementations the function must return the number of characters that
+    would have been printed had the buffer been large enough.  For conforming
+    vswprintf implementations the function must return a negative number
+    and set errno.
+
+    What vswprintf sets errno to is undefined but Darwin seems to set it to
+    EOVERFLOW.  The only expected errno are EILSEQ and EINVAL.  Both of
+    those are defined in the standard and backed up by several conformance
+    statements.  Note that ENOMEM mentioned in the manual page does not
+    apply to swprintf, only wprintf and fwprintf.
+
+    Official manual page:
+    http://www.opengroup.org/onlinepubs/009695399/functions/swprintf.html
+
+    Some conformance statements (AIX, Solaris):
+    http://www.opengroup.org/csq/view.mhtml?RID=ibm%2FSD1%2F3
+    http://www.theopengroup.org/csq/view.mhtml?norationale=1&noreferences=1&RID=Fujitsu%2FSE2%2F10
+
+    Since EILSEQ and EINVAL are rather common but EOVERFLOW is not and since
+    EILSEQ and EINVAL are specifically defined to mean the error is other than
+    an undersized buffer and no other errno are defined we treat those two
+    as meaning hard errors and everything else gets the old behavior which
+    is to keep looping and increasing buffer size until the function succeeds.
+
+    In practice it's impossible to determine before compilation which behavior
+    may be used.  The vswprintf function may have vsnprintf-like behavior or
+    vice-versa.  Behavior detected on one release can theoretically change
+    with an updated release.  Not to mention that configure testing for it
+    would require the test to be run on the host system, not the build system
+    which makes cross compilation difficult. Therefore, we make no assumptions
+    about behavior and try our best to handle every known case, including the
+    case where wxVsnprintf returns a negative number and fails to set errno.
+
+    There is yet one more non-standard implementation and that is our own.
+    Fortunately, that can be detected at compile-time.
+
+    On top of all that, ISO C99 explicitly defines snprintf to write a null
+    character to the last position of the specified buffer.  That would be at
+    at the given buffer size minus 1.  It is supposed to do this even if it
+    turns out that the buffer is sized too small.
+
+    Darwin (tested on 10.5) follows the C99 behavior exactly.
+
+    Glibc 2.6 almost follows the C99 behavior except vswprintf never sets
+    errno even when it fails.  However, it only seems to ever fail due
+    to an undersized buffer.
+*/

 #if wxUSE_UNICODE_UTF8
 template<typename BufferType>
 #else
 #if wxUSE_UNICODE_UTF8
 template<typename BufferType>
 #else


@@ -1609,12 +1678,19 @@ static int DoStringPrintfV(wxString& str,
         // only a copy
         va_list argptrcopy;
         wxVaCopy(argptrcopy, argptr);
         // only a copy
         va_list argptrcopy;
         wxVaCopy(argptrcopy, argptr);

+
+#ifndef __WXWINCE__
+        // Set errno to 0 to make it determinate if wxVsnprintf fails to set it.
+        errno = 0;
+#endif

         int len = wxVsnprintf(buf, size, format, argptrcopy);
         va_end(argptrcopy);
 
         // some implementations of vsnprintf() don't NUL terminate
         // the string if there is not enough space for it so
         // always do it manually
         int len = wxVsnprintf(buf, size, format, argptrcopy);
         va_end(argptrcopy);
 
         // some implementations of vsnprintf() don't NUL terminate
         // the string if there is not enough space for it so
         // always do it manually

+        // FIXME: This really seems to be the wrong and would be an off-by-one
+        // bug except the code above allocates an extra character.

         buf[size] = _T('\0');
 
         // vsnprintf() may return either -1 (traditional Unix behaviour) or the
         buf[size] = _T('\0');
 
         // vsnprintf() may return either -1 (traditional Unix behaviour) or the


@@ -1636,19 +1712,33 @@ static int DoStringPrintfV(wxString& str,
             // assume it only returns error if there is not enough space, but
             // as we don't know how much we need, double the current size of
             // the buffer
             // assume it only returns error if there is not enough space, but
             // as we don't know how much we need, double the current size of
             // the buffer

-            size *= 2;
+#ifndef __WXWINCE__
+            if( (errno == EILSEQ) || (errno == EINVAL) )
+            // If errno was set to one of the two well-known hard errors
+            // then fail immediately to avoid an infinite loop.
+                return -1;
+            else
+#endif // __WXWINCE__
+            // still not enough, as we don't know how much we need, double the
+            // current size of the buffer
+                size *= 2;

 #endif // wxUSE_WXVSNPRINTF/!wxUSE_WXVSNPRINTF
         }
         else if ( len >= size )
         {
 #if wxUSE_WXVSNPRINTF
 #endif // wxUSE_WXVSNPRINTF/!wxUSE_WXVSNPRINTF
         }
         else if ( len >= size )
         {
 #if wxUSE_WXVSNPRINTF

-            // we know that our own implementation of wxVsnprintf() returns 
+            // we know that our own implementation of wxVsnprintf() returns

             // size+1 when there's not enough space but that's not the size
             // of the required buffer!
             size *= 2;      // so we just double the current size of the buffer
 #else
             // some vsnprintf() implementations NUL-terminate the buffer and
             // some don't in len == size case, to be safe always add 1
             // size+1 when there's not enough space but that's not the size
             // of the required buffer!
             size *= 2;      // so we just double the current size of the buffer
 #else
             // some vsnprintf() implementations NUL-terminate the buffer and
             // some don't in len == size case, to be safe always add 1

+            // FIXME: I don't quite understand this comment.  The vsnprintf
+            // function is specifically defined to return the number of
+            // characters printed not including the null terminator.
+            // So OF COURSE you need to add 1 to get the right buffer size.
+            // The following line is definitely correct, no question.

             size = len + 1;
 #endif
         }
             size = len + 1;
 #endif
         }