Printing of the largest and other very large floats is wrong

[auroranockert]

If you run std::float::infinity.next_after(1.0)` you get the following (unexpected result)

=> 179769313486231680088648464220646842686668242844028646442228680066046004606080400844208228060084840044686866242482868202680268820402884062800406622428864666882406066422426822086680426404402040202424880224808280820888844286620802664406086660842040886824002682662666864246642840408646468824200860804260804068888.0

the expected result is actually something like

=> 179769313486231580788648464220646842686668242844028646442228680066046004606080400844208228060084840044686866242482868202680268820402884062800406622428864666882406066422426822086680426404402040202424880224808280820888844286620802664406086660842040886824002682662666864246642840408646468824200860804260804068887.0

[huonw]

Related to #6220.

[silene]

All the large floating-point numbers are multiple of 2^971, so the expected result should not be so odd (pun intended). The largest floating-point number is actually:

179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368

[metajack]

triage bump. i think one of these bugs should be production ready. nominating this one.

[catamorphism]

Accepted for production-ready

[pnkfelix]

With a heavy heart, not tagging as 1.0 blocker.

P-high, not 1.0.

[tshepang]

@pnkfelix why is this not a 1.0 blocker?

[pnkfelix]

@tshepang because at the time of that triage, we decided it would be acceptable to fix this after 1.0

As in, we were not saying we cannot fix it; just that we would not wait for perfection before releasing 1.0

I do hope the recently posted flt2dec PR gets into 1.0

[tshepang]

@pnkfelix what I wanted to know is why this was decided. For example, is the fix hard and/or time consuming?

[pnkfelix]

Yes, fixing this properly is hard. See PR #24612 for a taste of the difficulty