[patch, fortran] PR30096 - Fixes for entity ambiguity and interfaces (PR29975, PR30068)

[Paul Thomas]

Tobias,

:ADDPATCH fortran:

Thank you truly for the help on (i) the standards issues - your thread 
on comp.lang.fortran was extremely helpful in getting the advice of 
Richard Maine and Malcolm Cohen; and (ii) for doing the ChangeLogs.  I 
hope that this latter has given you some familiarity with the inner 
workings.  In addition, I enjoyed the collaboration and hope that we can 
do it again.

Just in case that I have screwed up this thread completely by modifying 
the title, here is the link to Tobias' submission.

 http://gcc.gnu.org/ml/fortran/2006-12/msg00124.html

In the course of a meeting yesterday, I realised how to fix PR30096, in 
which there was confusion between local and host-associated interfaces.  
The fix is a small extension of that for PR30068, so I have combined them.

Generally(always?), a local symbol has precedence over a host associated 
symbol, whether local to the host or use associated.  Thus, it is 
incorrect to compare host associated with the local version because they 
can never be ambiguous.  The fix was very simple and consisted of 
preventing interface.c(check_sym_interfaces) from checking  host 
associated interfaces with the same name - in this case, less is more 
:-)   The tescase, interface_9.f90 is heavily based on the reporter's 
test.  It has been made dg-run because the first version of the patch 
silenced the error but yielded the wrong procedure call. *sigh*

Regtested on Cygwin_NT/amd64 - OK for trunk and 4.2?

Paul



-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Change.Logs
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20061209/9703db42/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pr30068.diff
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20061209/9703db42/attachment-0001.ksh>