Ifort raises catastrophic error when compiling a moderatly errornous code

[jerzyg]

I'd like to report a catastrophic compiler error that occurs when compiling a slightly errornous code. The compiler reports "catastrophic error: **Internal compiler error: segmentation violation signal raised".

The problem appears when the "associate" construct is used to bind a name with a field of a derived-type variable.  It seems to be a necessary condition for this error to occur that the field of the variable must be an array. It is crucial that the definition of the derived type in question is not available at the compilation time (e.g. because of missing "use" or a wrong compilation order).

Please find attached reproducer.

I have observed this behavior with ifort 12.1.3 20120212 and 12.1.5 20120612, both on Linux 64bit (Intel64).

best regards,
Jerzy

[TimP]

ifort jerzy.f90
jerzy.f90(18): error #6404: This name does not have a type, and must have an explicit type.   [SOMEWHERE]
            associate (input => somewhere%member(1)%field)
--------------------------------^
jerzy.f90(18): error #6460: This is not a field name that is defined in the encompassing structure.   [MEMBER]
            associate (input => somewhere%member(1)%field)
....
My 12.1.5 throws the internal error at this point, but it looks like the next release will show additional diagnostics consequent to this and terminate normally.

[jerzyg]

Dear Tim,

thanks for your prompt reply. Indeed, Ifort 13.0.0.060 Beta Build 20120607 reacts in a correct way and gives much more diagnostic messages.

best regards,
Jerzy