Re: Bignum speedup patch causes crash at startup

[Pip Cet]

On Wed, Sep 5, 2018 at 3:41 PM Paul Eggert <address@hidden> wrote:
> or the eassert buried inside XFIXNUM

There is no eassert inside XFIXNUM, which makes it different from
XSYMBOL etc. I think there should be; that would break the code we're
talking about here, but IMHO it's much cleaner not to call XFIXNUM on
a non-fixnum, knowing that the clever way the code is written will
prevent the meaningless integer thus achieved from ever being used. I
doubt there is a performance gain to be had from that way of writing
code, and it's very hard to read. And there might even be a
performance win to assuming XFIXNUM's argument has the right tag
value.