Bug 113 – std.format.doFormat and small hex numbers

There's a bug in std.format.doFormat regarding small hexadecimals, specifically 0xA to 0xF. This small test program demonstrates the problem: ----- D:\Temp>cat test.d import std.stdio; int main() { for (int i = 10; i < 16; i++) writefln("%d - %x", i, i); return 0; } D:\Temp>dmd -run test.d 10 - : 11 - ; 12 - < 13 - = 14 - > 15 - ? ----- As you can see, this program doesn't produce the expected output (which would have the lines end in a-f instead of the various punctuation marks). Similar behavior can be observed by using std.string.format and probably anything else that relies on std.format.doFormat to perform such formatting. From a quick look at the code the bug seems to originate around line 875 of src/phobos/std/format.d where for single-character numbers (vnumber < base) the single-character output '0'+vnumber is produced. This "optimization" seems to result in this bug when vnumber > 9 and base > 10.

Ok, this definitely is a bug in the library. Currently the code looks like: if (vnumber < base) { vchar = '0' + vnumber; goto L2; } which results in a bug in case vnumber >= 10 and base > 10. To fix it, we could use: if (vnumber < base && vnumber < 10) { vchar = '0' + vnumber; goto L2; } else if (vnumber < base) { vchar = 'A' + (vnumber - 10); goto L2; }

(In reply to comment #1) > vchar = 'A' + (vnumber - 10); I think you mean something more like: vchar = (uc ? 'A' : 'a') + (vnumber - 10); but yeah, something like that. Or rewire the condition to just go to the normal code for (vnumber >= 10).

Fixed DMD 0.161 (thanks to Artem Rebrov)