Re: libnettle/libhogweed WIP

[Stefan Monnier]

> I think STRING_SET_CHARS is what you want here.

Really?  This sets ->size without changing ->size_bytes, so it doesn't
look right.  I think we don't have the function that Ted wants.
Basically, we'd need to provide a `resize_string_data` function, which
will have to massage the sdata blocks appropriately, paying attention to
LARGE_STRING_BYTES and things like that.
I think when shrinking the string, we can avoid copying in most cases,
but IIUC there are some cases where we can't avoid it (when the string
is smaller than LARGE_STRING_BYTES and we shrink it by too few bytes to
be able to place a new (free) sdata block in the freed space.
This said, it could get a bit ugly around the LARGE_STRING_BYTES
boundary because I can't see an easy way to efficiently figure out if
a string data was allocated into a large_sblock(s) or into a small block
(kept in oldest_sblock instead) so if the old size was larger than
LARGE_STRING_BYTES but the new size is smaller, we may have to add
a dummy "free sdata" as well (which again is only possible if the
shrinkage is sufficient to make room for such a dummy sdata)).


        Stefan