code review 100580044: cmd/gc: fix float32 const conversion and printing of bi...

cmd/gc: fix float32 const conversion and printing of big float consts

The float32 const conversion used to round to float64
and then use the hardware to round to float32.
Even though there was a range check before this
conversion, the double rounding introduced inaccuracy:
the round to float64 might round the value further away
from the float32 range, reaching a float64 value that
could not actually be rounded to float32. The hardware
appears to give us 0 in that case, but it is probably undefined.
Double rounding also meant that the wrong value might
be used for certain border cases.

Do the rounding the float32 ourselves, just as we already
did the rounding to float64. This makes the conversion
precise and also makes the conversion match the range check.

Finally, add some code to print very large (bigger than float64)
floating point constants in decimal floating point notation instead
of falling back to the precise but human-unreadable binary floating
point notation.

Fixes issue 8015.

[rsc, 2014-05-20 00:35:10]

Hello golang-codereviews@googlegroups.com (cc: iant, r),

I'd like you to review this change to
https://code.google.com/p/go/

[iant, 2014-05-20 01:03:28]

LGTM

Your tests pass with gccgo modolu error message syntax, which is encouraging.

[rsc, 2014-05-20 02:58:01]

*** Submitted as https://code.google.com/p/go/source/detail?r=5b9ac653acf6 ***

cmd/gc: fix float32 const conversion and printing of big float consts

The float32 const conversion used to round to float64
and then use the hardware to round to float32.
Even though there was a range check before this
conversion, the double rounding introduced inaccuracy:
the round to float64 might round the value further away
from the float32 range, reaching a float64 value that
could not actually be rounded to float32. The hardware
appears to give us 0 in that case, but it is probably undefined.
Double rounding also meant that the wrong value might
be used for certain border cases.

Do the rounding the float32 ourselves, just as we already
did the rounding to float64. This makes the conversion
precise and also makes the conversion match the range check.

Finally, add some code to print very large (bigger than float64)
floating point constants in decimal floating point notation instead
of falling back to the precise but human-unreadable binary floating
point notation.

Fixes issue 8015.

LGTM=iant
R=golang-codereviews, iant
CC=golang-codereviews, r
https://codereview.appspot.com/100580044

[gobot, 2014-05-20 03:06:15]

This CL appears to have broken the freebsd-386 builder.
See http://build.golang.org/log/3c79e2466c4282e725d8aa0cd2efce7234ba8fed

[lucio, 2014-08-07 10:59:19]

It seems new arguments were added to the print invocation in
cmd/gc/mparith1.c:594 without adding corresponding format keywords.  The first
of these (exp) is an integer, so it's easy to fix, but the second one is a
multiple precision value and I'm not sure what the corresponding format might
be.

The plan 9 compilers report a warning.

Lucio.

[rsc, 2014-08-07 12:07:30]

Thanks. See http://codereview.appspot.com/123880043.