Bug 1113793

That should be:

 % make -f Makefile.intel zpipe.o 

To see what ifort reports as the syntax error.

Created attachment 912913 [details]
Simplified test case exhibiting seg-fault

This is an explicit test case which causes the segmentation fault in gfortran as described in bug 1113793. It can be run with no additional flags (i.e. `gfortran test.f`).

Comment on attachment 912913 [details]
Simplified test case exhibiting seg-fault

The segmentation fault occurrs when an old-style initialization is present within a derived type declaration. 

The internal function which handles old-style initializations as part of a variable declaration fails to handle the case when the variable being declared is a derived type component.

Created attachment 912938 [details]
Quick fix to avoid seg-faulting in this case

This is a simple patch for gfortran v4.8.3 (and likely other versions) which throws a syntax error when old-style initializations are used in a derived-type component declaration (thereby avoiding a segmentation fault).

This behavior is justified by the behavior of ifort v11.1+ (the Intel FORTRAN compiler). The Intel compiler also throws a syntax error for attachment 912913 [details] (test.f).

If old-style initializations should be parsed and resolved like standard initializations, the feature should probably be enabled only as a GNU extension. Some additional work may have to be done in gcc/fortran/decl.c, either in match_old_style_init() or variable_decl().


The patch can be applied with the following command from the top-level source directory (e.g. /..../src/gcc-4.8.3/):

patch -p0 < b1113793.patch

I think it is fine to reject it, though with your patch you'll leak as memory in the caller.  Here is an alternative (so far untested) patch:

2014-07-01  Jakub Jelinek  <jakub>
	    Fritz Reese  <Reese-Fritz>

	* decl.c (variable_decl): Reject old style initialization
	for derived type components.

	* gfortran.dg/oldstyle_5.f: New test.

--- gcc/fortran/decl.c.jj	2014-06-30 09:28:50.000000000 +0200
+++ gcc/fortran/decl.c	2014-07-01 16:47:19.466050044 +0200
@@ -1997,6 +1997,13 @@ variable_decl (int elem)
       if (!gfc_notify_std (GFC_STD_GNU, "Old-style "
 			   "initialization at %C"))
 	return MATCH_ERROR;
+      else if (gfc_current_state () == COMP_DERIVED)
+	{
+	  gfc_error ("Invalid old style initialization for derived type "
+		     "component at %C");
+	  m = MATCH_ERROR;
+	  goto cleanup;
+	}
 
       return match_old_style_init (name);
     }
--- gcc/testsuite/gfortran.dg/oldstyle_5.f.jj	2014-07-01 16:50:40.449001427 +0200
+++ gcc/testsuite/gfortran.dg/oldstyle_5.f	2014-07-01 16:48:33.000000000 +0200
@@ -0,0 +1,8 @@
+C { dg-do compile }
+      TYPE T
+      INTEGER A(2)/1,2/ ! { dg-error "Invalid old style initialization for derived type component" }
+      END TYPE
+      TYPE S
+      INTEGER B/1/ ! { dg-error "Invalid old style initialization for derived type component" }
+      END TYPE
+      END

Committed: http://gcc.gnu.org/r212227 , http://gcc.gnu.org/r212228 , http://gcc.gnu.org/r212229 .  Backportable to 4.4-RH.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

http://rhn.redhat.com/errata/RHBA-2014-1377.html