Re: Bignum speedup patch causes crash at startup

[Paul Eggert]

Eli Zaretskii wrote:
> > > C:/emacs/git/emacs/master/src/data.c:2962:12: note: 'accum' was declared here
> > >      intmax_t accum = XFIXNUM (val);
> > GCC complains that 'accum' might be used uninitialized even though 'accum's
> > declaration has an initializer? Looks like a clear GCC bug, and might be worth
> > reporting there.
> Maybe it's because eassert can abort?

Possibly, if GCC is buggy. But if GCC is working properly, it can easily tell 
that the preceding eassume (or the eassert buried inside XFIXNUM) cannot 
possibly cause control flow to jump over the initialization and into a use of 
'accum', because any signal must jump out of that block and therefore cannot use 
'accum'.