[Patch, fortran] [4.4/4.5/4.6 regression] ICE in resolve_equivalence

[Mikael Morin]

Hello,

this patch fixes pr44660 where a symbol pointer is kept in the gfc_equiv 
list even if the statement is rejected and the symbol freed.
The patch restores the previous equivalence list if the statement is 
rejected.

There is no testcase as the ones in the pr were not giving any ICE on my 
platform. I can add them if someone prefers. I have checked that the 
valgrind errors are gone for both of them anyway.

Regression tested on x86_64-unknown-freebsd8.0. OK for trunk/4.5/4.4 ?
I need release manager approval for accelerated commit on 4.5 (if not 
too late).

Mikael.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pr44660.log
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20100725/439f9f06/attachment.log>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pr44660.diff
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20100725/439f9f06/attachment.ksh>