http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54917
Bug #: 54917
Summary: transfer on polymorphic variable causes ICE
(gfc_target_expr_size)
Classification: Unclassified
Product: gcc
Version: 4.7.3
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
AssignedTo: unassig...@gcc.gnu.org
ReportedBy: quanthe...@gmail.com
Created attachment 28436
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28436
Short program that causes the bug
When compiling a routine that uses transfer on a polymorphic variable, the
following command fails:
gfortran -c -Wsurprising test_module.F90
The error message is:
test_module.F90:18.22:
end module test_module
1
Internal Error at (1):
Invalid expression in gfc_target_expr_size.
This error does not occur if -Wsurprising is off and the polymorphic variable
is the "source". It does occur if the polymorphic variable is the "mold", even
with no warnings on.
Obviously, you would rarely-or-never want to actually do such a transfer, but I
did notice this problem because I have been using the following macro to turn
off warnings about unused arguments:
#define UNUSED_VAR(arg) if (.false.) write(*,*) transfer(arg,1)
A bit hack-ish, but gets the job done in impure routines. In any case, even if
gfortran rejects this, getting a specific error message rather than an internal
error would be nice.
I'm in a somewhat limited environment right now, so I can't tell for sure
whether this is an old problem or a regression. This has happened with one of
the 4.7.3 unofficial prerelease builds:
gcc-4.7 -v
Using built-in specs.
COLLECT_GCC=gcc-4.7
COLLECT_LTO_WRAPPER=/home/santos/gcc-4.7/bin/../libexec/gcc/x86_64-unknown-linux-gnu/4.7.3/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-4.7-source/gcc-4.7-20120929/configure
--enable-languages=c,c++,fortran --enable-checking=release --disable-bootstrap
--disable-libmudflap --enable-libgomp --enable-lto --enable-gold
--with-plugin-ld=/usr/bin/gold --prefix=/usr/local/gcc-4.7
Thread model: posix
