[m-rev.] diff: workaround seg. fault in srcdist on 64-bit machines

[Julien Fischer]

On Tue, 29 Jun 2010, Zoltan Somogyi wrote:

> On 29-Jun-2010, Julien Fischer <juliensf at csse.unimelb.edu.au> wrote:
>> A proper fix would be to disable the generation of non-portable code for
>> switches when --cross-compiling is enabled.
>
> That should be trivial to do: add an option implication from cross_compiling to
> not smart_indexing in handle_options.m. There are already many other such
> implications there.

Disabling smart indexing entirely when --cross-compiling is enabled is
overkill IMO -- however I don't think we currently have options that
would allow finer grained control of smart indexing.

> However, we should definitely find out what the actual problem is, since
> this may not be it.

I agree and I intend to look into this further when I have the time.
(I would guess it is the parts of switch generation that call
int.bits_per_int/0 since they don't appear to be protected by
--cross-compiling; it was by grepping the compiler for bits_per_int and
then looking at whether the value of --cross-compiling was checked that
gave me the idea of turning it off.)

Julien.
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------