Bad variable initialization in block causes ICE

[IanH]

Sometimes I forget what is permitted and what is not.  This is non-conforming code (says F2008 7.1.12p2), but ifort perhaps got more confused than it should.

PROGRAM MentalBlock
  IMPLICIT NONE
  BLOCK
    INTEGER :: array(2) = SIZE(array)
  END BLOCK
END PROGRAM MentalBlock

>ifort /check:all /warn:all /standard-semantics "2016-02-07 MentalBlock.f90"
Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 16.0 Build 20151021
Copyright (C) 1985-2015 Intel Corporation.  All rights reserved.

101004_2042

catastrophic error: **Internal compiler error: internal abort** Please report this error along with the circumstances in
 which it occurred in a Software Problem Report.  Note: File and line given may not be explicit cause of this error.
compilation aborted for 2016-02-07 MentalBlock.f90 (code 1)

A variant, that perhaps is conforming...

PROGRAM MentalBlockDeluxe
  IMPLICIT NONE
  BLOCK
    DIMENSION :: array(2)
    INTEGER :: array = SIZE(array)
  END BLOCK
END PROGRAM MentalBlockDeluxe

>ifort /check:all /warn:all /standard-semantics "2016-02-07 MentalBlockDeluxe.f90"
Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 16.0 Build 20151021
Copyright (C) 1985-2015 Intel Corporation.  All rights reserved.

101004_2042

catastrophic error: **Internal compiler error: internal abort** Please report this error along with the circumstances in
 which it occurred in a Software Problem Report.  Note: File and line given may not be explicit cause of this error.
compilation aborted for 2016-02-07 MentalBlockDeluxe.f90 (code 1)

.

[Steven_L_Intel1]

Thanks - I will send this off to the developers.

[Steven_L_Intel1]

Escalated as issue DPD200381580. Interestingly, if the declaration is not within a BLOCK the compiler accepts it, but shouldn't. 

[Steven_L_Intel1]

A further update. This usage is not standard F2008 but will be standard in F2015 (paper 15-194), which carved out an exemption for initialization expressions. I've asked our developers to make sure that standards warnings are handled appropriately for this.

[Steven_L_Intel1]

The ICE will be fixed in the final 17.0 release later this year.