Re: EXPR_WITH_FILE_LOCATION Tree Nodes

[Ranjit Mathew <rmathew at gmail dot com>]

On Tue, 08 Feb 2005 23:13:00 -0800, Per Bothner <per@bothner.com> wrote:
> Ranjit Mathew wrote:
> > Hi,
> >
> >   While parsing Java source files, the current front-end
> > creates EWFL wrapper nodes for identifiers, etc. Do I
> > understand correctly that every statement/expression must
> > have a wrapper EWFL node?
> 
> No, only those for while line number information is desired.

But then we already have EXPR_WFL_LINECOL for this,
no? (Which works for all tree nodes as you point out.)


> And in any case, EXPR_WITH_FILE_LOCATION should go away.
> It won't be needed with the new USE_MAPPED_LOCATION code

Slight digression: would USE_MAPPED_LOCATION be
the default in 4.1? (I seem to remember that only the Ada
front-end remained to be converted.)


> - except that EXPR_WITH_FILE_LOCATION is used in some rather
> complex ways.  But hopefully gcjx will make it moot.

gcjx will hopefully make a *lot* of things moot! :-)


> > If not, why does parse_error_context() expect an EWFL
> > always and barfs when it doesn't get it? (See PR8608, for
> > example.)
> 
> Who knows.  EXPR_WITH_FILE_LOCATION is used in some rather
> strange ways.

In that case, would it be acceptable to make
parse_error_context() and issue_warning_error_from_context()
handle both EWFL and "normal" nodes?

That should at least help in cases like PR8608.

Thanks for answering my queries,
Ranjit.

-- 
Ranjit Mathew      Email: rmathew AT gmail DOT com

Bangalore, INDIA.    Web: http://ranjitmathew.hostingzero.com/