Segfault when using finalizer

[Andrew Benson]

I opened PR 96122 for the following:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96122

--

The following code segfaults with gfortran 11.0.0:


module a
  type vs
     character(len=1), dimension(:), allocatable :: ch
   contains
     final :: vsf
  end type vs
  type mtd
     type(vs), dimension(2) :: n
  end type mtd
contains
  subroutine c()
    type(mtd) :: m
    allocate(m%n(1)%ch(1))
    m%n(1)%ch(1)="a"
    allocate(m%n(2)%ch(1))
    m%n(2)%ch(1)="b"
    return
  end subroutine c
  subroutine vsf(s)
    type(vs), intent(inout) :: s
    if (allocated(s%ch)) deallocate(s%ch)
    return
  end subroutine vsf
end module a
program b
  use a
  call c()
end program b



$ gfortran -v
Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/home/abenson/Galacticus/Tools/libexec/gcc/x86_64-pc-
linux-gnu/11.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-git/configure --prefix=/home/abenson/Galacticus/Tools --
enable-languages=c,c++,fortran --disable-multilib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 11.0.0 20200705 (experimental) (GCC) 

$ gfortran tmp.F90 -g

$ ./a.out 

Program received signal SIGSEGV: Segmentation fault - invalid memory 
reference.

Backtrace for this error:
#0  0x7f47668d61ef in ???
        at /data001/abenson/Galacticus/Tools/glibc-2.12.1/signal/../sysdeps/
unix/sysv/linux/x86_64/sigaction.c:0
#1  0x40118f in __a_MOD___final_a_Vs
        at /home/abenson/Galacticus/Nightlies/galacticus/tmp.F90:30
#2  0x400bfa in __a_MOD___final_a_Mtd
        at /home/abenson/Galacticus/Nightlies/galacticus/tmp.F90:30
#3  0x401556 in __a_MOD_c
        at /home/abenson/Galacticus/Nightlies/galacticus/tmp.F90:22
#4  0x401562 in b
        at /home/abenson/Galacticus/Nightlies/galacticus/tmp.F90:34
#5  0x401599 in main
        at /home/abenson/Galacticus/Nightlies/galacticus/tmp.F90:33
Segmentation fault


The problem doesn't occur if the finalizer for type "vs" isn't used. It also 
doesn't occur if the "n" component of type "mtd" has dimension(1) or is a 
scalar.

The finalizer in this case is pointless (since it just deallocates an 
allocatable, which would happen automatically if no finalizer were used), but 
as far as I can see is legal code.

This appears to be a regression (worked in 10.1) but I haven't yet tracked 
down exactly where the problem was introduced.
-- 

* Andrew Benson: http://users.obs.carnegiescience.edu/abenson/contact.html

* Galacticus: https://github.com/galacticusorg/galacticus