This adds some more test cases about the bug that commit
95ea86d5 fixes.
The bug was that in the perl 5.12 branch, sprintf("%.2f") returns ""
whereas it should return "0.00" -- but this only applies in the very
special cases, not eg. when the template has extra characters. The
log message and the tests of that commit only talks about "%.0f", but
I think it would be worth to also test the "%.2f".
>%f< >0< >0.000000<
>%.0f< >[]< >0 MISSING<
> %.0f< >[]< > 0 MISSING<
+>%.2f< >[]< >0.00 MISSING<
+>%.2fC< >[]< >0.00C MISSING<
>%.0f< >0< >0<
>%.0f< >2**38< >274877906944< >Should have exact int'l rep'n<
>%.0f< >0.1< >0<
>%#g< >12345.6789< >12345.7<
>%.0g< >[]< >0 MISSING<
> %.0g< >[]< > 0 MISSING<
+>%.2g< >[]< >0 MISSING<
+>%.2gC< >[]< >0C MISSING<
>%.0g< >-0.0< >-0< >C99 standard mandates minus sign but C89 does not skip: MSWin32 VMS hpux:10.20 openbsd netbsd:1.5 irix darwin<
>%.0g< >12345.6789< >1e+04<
>%#.0g< >12345.6789< >1.e+04<