Bug 3251 – DecimalFloat literal cannot begin with "08" or "09"

DecimalFloat: DecimalDigits . DecimalDigits . DecimalDigits DecimalDigits . DecimalDigits DecimalExponent . Decimal . Decimal DecimalExponent DecimalDigits DecimalExponent DecimalDigits: DecimalDigit DecimalDigit DecimalDigits DecimalDigit: 0 NonZeroDigit _ Going by this, floating point literals are the same whether they begin with a 0 or not, unlike integer literals. Code that illustrates this fact: ---------- pragma(msg, (76).stringof); pragma(msg, (076).stringof); pragma(msg, (76.).stringof); pragma(msg, (076.).stringof); pragma(msg, (76.543).stringof); pragma(msg, (076.543).stringof); pragma(msg, (678.).stringof); pragma(msg, (0678.).stringof); pragma(msg, (876.543).stringof); //pragma(msg, (0876.543).stringof); ---------- 76 62 76 76 76.543 76.543 678 678 876.543 ---------- However, if the final, commented-out line is reinstated, I get errors: ---------- float_literal.d(10): found '876.543' when expecting ')' float_literal.d(10): no identifier for declarator .stringof float_literal.d(10): semicolon expected, not ')' float_literal.d(10): Declaration expected, not ')' ---------- Presumably, DMD sees the '0' and begins trying to lex it as an Octal, but fails to recover properly when it stumbles upon the '8'. But it's strange that other similar forms compile successfully, such as the aforementioned "0678.", "008." and "0_9.".

Currently, when dmd sees a 0 it starts parsing an octal. If it finds any digit > 7 it errors. However, if it finds a point, it immediately switches to parsing a floating point literal, thus accepting from this point digits larger than 7: 07341 => octal 07238 => error 024.8 => foating point literal I think this is fine.