printf: Avoid setting FE_INEXACT in printf()

[arichardson]

Summary

The k = (int)ds; cast inside gdtoa/dtoa can set FE_INEXACT.
Having printf() change the FE_INEXACT flag results in some of the
msun tests failing when compiling them with debugging flags.
To avoid this, save the state of FE_INEXACT before calling gdtoa()
and clear it if gdtoa() set it.

Test Plan

building msun with -DDOPRINT -DDEBUG -DDOPRINT_LD80 no longer causes all tests to fail

[dim]

Hm, wouldn't it be better to fix this in gdtoa() instead? Then it would be solved for all gdtoa() consumers. That said, I've understood that the gdtoa code is very impenetrable...

[arichardson]

In D28787#646949, @dim wrote:

Hm, wouldn't it be better to fix this in gdtoa() instead? Then it would be solved for all gdtoa() consumers. That said, I've understood that the gdtoa code is very impenetrable...

I had a brief look at modifying the gdtoa code, but it seemed to be quite akward so I chose to change the wrapper instead.

[dim]

In D28787#646959, @arichardson wrote:

In D28787#646949, @dim wrote:

Hm, wouldn't it be better to fix this in gdtoa() instead? Then it would be solved for all gdtoa() consumers. That said, I've understood that the gdtoa code is very impenetrable...

I had a brief look at modifying the gdtoa code, but it seemed to be quite akward so I chose to change the wrapper instead.

Well, if printf is the only consumer of gdtoa, then this is OK with me. Otherwise I'd just hide the "real" gdtoa behind the wrapper for everybody.

[arichardson]

• fix softfloat build

[arichardson]

fix typo

[dim]

If this also works for softfloat now, please go ahead.