Re: [Emacs-diffs] master c8d14cf: Fix glitches in recent hash table changes

[Paul Eggert]

On 02/19/2017 12:29 PM, Stefan Monnier wrote:
> I very consciously decided that those rounding errors do not matter.

Yes, the exact values of hash-table size parameters do not matter. 
Still, it's surprising to see that seemingly-spurious change from 0.8 to 
0.800000011920929. I fixed this by changing the default to 0.8125, which 
does not suffer rounding error.

Also, we can tweak rehash_size and next_hash similarly. I installed a 
set of patches to do that. On my x86-64 Fedora 25 platform the patches 
save 8 bytes per hash table and a few hundred bytes in executable code. 
No big deal, but the code seems a bit simpler this way.