Re: allocate_string_data memory corruption

[Stefan Monnier]

>     For the sake of the release, how about just moving the threatened
>     portion of allocate_vectorlike into the BLOCK_INPUT section that's
>     already present in the function?

> Ok.

There's no BLOCK_INPUT used in allocate_vectorlike unless DOUG_LEA_MALLOC
is defined.  But admittedly, allocate_vectorlike is very slow anyway (at
least on platforms where we use conservative stack scanning, i.e. in 99% of
the cases) so adding a BLOCK_INPUT there isn't a problem.

> But can all the known problems be fixed this way?

At the cost of adding BLOCK_INPUT for each allocation function?  Sure, but
I'd rather not do that.


        Stefan