Re: [patch] mem_node shrink

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> So I think MOST_POSITIVE_FIXNUM should be replaced with something
>> guaranteed to be right.  Define a constant to serve as the width of
>> that field, and use the same constant here in something like -((-1) <<
>> MEM_NODE_SIZE_WIDTH).

>     If GC_MALLOC_CHECK is defined, xmalloc and xrealloc calls
>     mem_insert.  On 64-bit system, if neither USE_MMAP_FOR_BUFFERS nor
>     REL_ALLOC are used, xmalloc and xrealloc may grow buffer text up
>     to the size which fits into integer, but not into 28-bit bitfield.
>     So, the bitfield should be wide enough, i.e. have width
>     BITS_PER_EMACS_INT - 4.

> There are two questions here.  You're talking about whether the field
> is wide enough.  The other is how to write this error check:

>> +  if (size > MOST_POSITIVE_FIXNUM)
>> +    abort ();

Yes.  A simpler test is to read the size we just wrote in the object's
field and check it reads out as the same value.


        Stefan