[cfe-commits] r56051 - /cfe/trunk/lib/CodeGen/CGCall.cpp

[Daniel Dunbar]

Yes, then intent is certainly to do this in a target specific manner. As
my comment said, since we weren't ever trying to be ABI complaint for
any platform my assumption was this would not break things people
were actually doing.

If this is causing problems for you I will go ahead and do it in a simple
target specific manner. It might be Wed. before I get to it, however.

 - Daniel

On Mon, Oct 6, 2008 at 8:28 AM, Matthijs Kooijman <> wrote:

> Hi Daniel,
>
> >   - As a test, enable basic usage for some common x86-32 cases. This
> >     increases our x86-32 compliance (on other targets our compliance
> >     will just be broken in a different way).
> It seems that this coercion happening now is only correct for x86? For our
> (custom) backend, this is not wanted and even breaks (because this
> conversion
> of structs to integer return types confuses scalarrepl later on, after
> inlining).
>
> Would it make sense to do this in more per-backend manner, or at least
> allow
> the TargetInfo to disable this?
>
> Gr.
>
> Matthijs
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
>
> iD8DBQFI6i6hz0nQ5oovr7wRAqQ4AJ4h5I/FbnP1Ca+gr5ZWtbuNAzrqegCfY82F
> mzUEqm/yhSWbMDdQ/lUHwOM=
> =9ozM
> -----END PGP SIGNATURE-----
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20081006/9d42c8e0/attachment.html>