[tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[Jan Hubicka]

Hi,
filter_expr and exc_ptr_expr can appear at left hand side of
modify_expr.
According to the current GIMPLE grammar this is invalid.  It seems to me
that these should be accepted as gimple ID, is that right?

Honza

Re: [tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[Richard Henderson]

On Sat, Dec 13, 2003 at 01:09:07AM +0100, Jan Hubicka wrote:
> According to the current GIMPLE grammar this is invalid.  It seems to me
> that these should be accepted as gimple ID, is that right?

Yes.

Really we should be using magic decls, but that means moving quite
a lot of except.c into tree-except.c; something I'm not prepared to
do right away.


r~

Re: [tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[Jan Hubicka]

> On Sat, Dec 13, 2003 at 01:09:07AM +0100, Jan Hubicka wrote:
> > According to the current GIMPLE grammar this is invalid.  It seems to me
> > that these should be accepted as gimple ID, is that right?
> 
> Yes.
> 
> Really we should be using magic decls, but that means moving quite
> a lot of except.c into tree-except.c; something I'm not prepared to
> do right away.

OK, thanks.
I've bundled this into the verify_gimple_grammar patch already, so
everything should be fine.

Honza
> 
> 
> r~

Re: [tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[law]

In message <20031216204323.GE20094@atrey.karlin.mff.cuni.cz>, Jan Hubicka write
s:
 >> On Sat, Dec 13, 2003 at 01:09:07AM +0100, Jan Hubicka wrote:
 >> > According to the current GIMPLE grammar this is invalid.  It seems to me
 >> > that these should be accepted as gimple ID, is that right?
 >> 
 >> Yes.
 >> 
 >> Really we should be using magic decls, but that means moving quite
 >> a lot of except.c into tree-except.c; something I'm not prepared to
 >> do right away.
 >
 >OK, thanks.
 >I've bundled this into the verify_gimple_grammar patch already, so
 >everything should be fine.
Any chance you could verify that tree-simple.[ch] comments are up-to-date in
regards to this extension to the simple grammar?  

jeff

Re: [tree-ssa] filter_expr and exc_ptr_expr as gimple ID?

[Jan Hubicka]

> In message <20031216204323.GE20094@atrey.karlin.mff.cuni.cz>, Jan Hubicka write
> s:
>  >> On Sat, Dec 13, 2003 at 01:09:07AM +0100, Jan Hubicka wrote:
>  >> > According to the current GIMPLE grammar this is invalid.  It seems to me
>  >> > that these should be accepted as gimple ID, is that right?
>  >> 
>  >> Yes.
>  >> 
>  >> Really we should be using magic decls, but that means moving quite
>  >> a lot of except.c into tree-except.c; something I'm not prepared to
>  >> do right away.
>  >
>  >OK, thanks.
>  >I've bundled this into the verify_gimple_grammar patch already, so
>  >everything should be fine.
> Any chance you could verify that tree-simple.[ch] comments are up-to-date in
> regards to this extension to the simple grammar?  

There is no specification of "ID" in the grammer, so there is no real
place to update.

I was thinking about a strategy to the grammar overall.  Currently
tree-simple.[ch] is out of date in several interesting ways, so I think
best thing to do is to get verify_gimple_grammar implementation I sent
last week into acceptable shape and then I can take a job of ensuring
that the implementation of verifier match the comments in future.
Having verifier will hopefully make the grammar more clean.

Honza
> 
> jeff