is it a c++ number ?

[jj2007]

Thanks. "For C and C++, octal numbers (base 8 ) begins with a zero" sounds like a can of worms, since "0123" is a valid decimal, too, and octal numbers are very rare nowadays 8)

[TouEnMasm]

I've seen that your above code must reworked because it:
- does reject valid FP numbers starting with a dot (e.g. ".123")
- accepts invalid octal numbers like "0123456789"
- does not recognized the "u" for MSVC integer suffixes (e.g. "uI32")
- accepts "0b", which is invalid
...

corrected in my code sample
Yours had some problems with real at end of list

http://masm32.com/board/index.php?topic=3601.msg37912#msg37912

[qWord]

Yours had some problems with real at end of list

which one exactly? "-1.18E - 38" is not a valid literal (the compiler recognize that as "-_unary 1.18E_invalid -_operator38_int")

EDIT: just some more stuff you latest code wrongly interpret:
- "0x" is accepted
- "0x.0p0" is not accepted (as said, when targeting C++ this format is not required)
- "0x123fff.ffae+23" is accepted
- suffix "i9" for integers is accepted
- ".9u" is accepted (invalid FP suffix)
- ".E-1" is accepted (at least on digit is required for the mantissa)

[TouEnMasm]

The "-1.18E - 38" is a limit and a valid littteral.
the strtod function return -1.18000 as value.

try:

      x = strtod( string, &stopstring );
      printf( "string = %s\n", string );
      printf("   strtod = %f\n", x );
      printf("   Stopped scan at: %s\n\n", stopstring );

and you get:

string = -1.18E - 38
   strtod = -1.180000
   Stopped scan at: E - 38

The i9 is reject by the compiler because he don't know a 9 bits register,I had only take care of the form.

[qWord]

The "-1.18E - 38" is a limit and a valid littteral.

not sure what you are trying to archive, but  yes, until "E" is valid. However, you will not fined any C or C++ compiler that accepts "-1.18E - 38".

Just as side note, after looking into the C99 standard, I've recognized that these hexadecimal floating point literals must always have a exponent: .e.g. "0x1.0p0" (just "0x1.0" would be invalid). Even this rule makes sense, because otherwise the f-suffix (float) would become ambiguous.

[qWord]

The attachment contains a recent adaptation of some old code that I use to verify that an input string, that is supposed to represent a double, is correctly formatted.
[..]

Code Select Expand

    static char pats[18][7] = {"d","pd","sd","ded","dpd","spd","desd",
                               "pded","sded","sdpd","dpded","pdesd",
                               "sdesd","spded","dpdesd","spdesd",
                               "sdpded","sdpdesd"};

Really nice idea! Looks a bit like regular expressions

[TouEnMasm]

However, you will not fined any C or C++ compiler that accepts "-1.18E - 38".

No need to search,it is vc express edition 2010.

[qWord]

However, you will not fined any C or C++ compiler that accepts "-1.18E - 38".

No need to search,it is vc express edition 2010.

I thought you want validate input from a c++ (or c) source file...