[Patch, Fortran] F2003: More ASSOCIATE stuff

[Daniel Kraft]

Hi all,

here's another ASSOCIATE patch.  It shifts handling of the variables and 
their initialization to the trans-* phase.  This fixes the problems with 
array boundaries and implements association to variables.  I think most 
of the useful stuff should work with this now.

What does not work (at least not in the complicated cases, like the one 
commented out in associate_1.f03 for now) is association to strings.  I 
could not figure this out, and would like to get the basic handling in 
this patch in to start cleanly with the remaining other bits :)

What also does not work is parsing of component references if you 
associate a variable to a derived-type value; the parser does not yet 
know that it should expect something like '%comp' following this name, 
and unfortunately the type of the target expression and thus of the 
associate-name is not known until resolution in certain cases (I think) 
-- and I do not see a solution at the moment, how we can still parse the 
component references cleanly before resolution without substantial 
reworks (like resolving the variables before parsing the executable 
statements -- which should work in theory?).  Any ideas?  But for now, 
I'd also like to leave this open.

And finally, names associated to expressions may not appear in "variable 
definition contexts"; currently, this is diagnosed for things like "a = 
5" but misses a lot of other cases (i.e., passing to INTENT([IN]OUT) 
dummys).  I do not think we currently have a general implementation of 
"variable definition context" as defined in the standard; do we?  Maybe 
this would be worthwhile to implement as a follow-up patch, too -- and 
in general, not just special-coding the checks for associate names. 
What do you think there?

But for now, the test was regtested on GNU/Linux-x86-32.  I saw two 
failures, namely array_memcpy_3.f90 and bind_c_dts_3.f90 -- but don't 
see how those can be related to ASSOCIATE.  Is this currently broken? 
Or maybe something changed recently and I got caught in a partial svn 
update?  (Although I don't think so.)

Ok for trunk when this is figured out?

Yours,
Daniel

-- 
http://www.pro-vegan.info/
--
Done:  Arc-Bar-Cav-Ran-Rog-Sam-Tou-Val-Wiz
To go: Hea-Kni-Mon-Pri
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch.changelog
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20100816/eca0357e/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch.diff
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20100816/eca0357e/attachment-0001.ksh>