RE: RE: varasm fix for

["Hartmut Penner" <HPENNER at de dot ibm dot com>]

>Hartmut Penner <hpenner@tonto.boeblingen.de.ibm.com> writes:

> > This fix is prerequisit for adding the s390 backend to gcc-3.0.
General, the
> > add of s390 backend into gcc-3.0 is approved by Mark Mitchell. Also the
fix
> > down below was approved with editoral changes to be done.

> Is this fix needed for the trunk also?

I would be good to have that fix also in the trunk.
But Mark Mitchell was writing me this:

>   3. What would be really, really useful would be to break all the
>     constant pool handling out into a new file, with a relatively
>     compact external interface (basically `get_pool_constant' and
>     such, with some documentation about how all this stuff is
>     supposed to work.  Comments for `struct pool_constant' would
>     be a start, at least.  Sames goes for `struct rtx_const'.  In
>     general, I'd give this varasm.c code about a C+, if I were
>     grading it.

>     You don't really have to do this, as part of your patch, and
>     we wouldn't want it for the 3.0 branch anyhow, but if, when
>     you figure out how something works, you clean up the code there,
>     that's a good thing.

> Andreas

> > Please tell me, if it is ok to commit.
> >
> > 2000-07-30  Hartmut Penner <hpenner@de.ibm.com>
> >  * varasm.c (decode_rtx_const): Allow unspec (symbol_ref) in constant
> >  pool to be identical by string address and index.

--

Mit freundlichem Gruß / Best regards,
                                 Hartmut Penner