printf with %g width specifier work abnormal in certain situation

[lc4]

printf("(%10g)\n", 100000.0f);  // --> (100000)
printf("(%10g)\n", 100001.0f);  // --> (    100001)
printf("(%10g)\n", 99999.0f);   // --> (     99999)

The first line should output: ( 100000) instead.

Tested on version 2.3.0 and 2.4.0, same result.

[skeeto]

Yup, this appears to be a very old bug (i.e. present in every w64dk) in the upstream Mingw-w64 __pformat_gfloat (mingw_pformat.c). I can confirm it with: $ cc -D__USE_MINGW_ANSI_STDIO=1 main.c $ ./a (100000) ( 100001) $ cc -D__USE_MINGW_ANSI_STDIO=0 main.c $ ./a ( 100000) ( 100001) Where the latter uses msvcrt!printf instead of Mingw-w64's implementation. I suspect it's related to the line: stream->width += stream->precision; It's formatting the number with the default precision, making decisions about the field width, changing the precision, then formatting again under the new configuration, which then messes up the computed field width. It would be worth opening a Mingw-w64 bug report for this.