[Bug fortran/26155] p cos(alat(j)) command make gdb Aborted (core dumped)

["vries at gcc dot gnu.org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26155

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
                 CC|                            |andrew.burgess at embecosm dot com
                   |                            |, vries at gcc dot gnu.org
   Last reconfirmed|                            |2021-02-03
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
Reproduced:
...
$ gfortran test.F90 -g -O0
$ gdb -batch a.out -ex start -ex "p cos(10)"
Temporary breakpoint 1 at 0x400b3d: file test.F90, line 31.

Temporary breakpoint 1, MAIN__ () at test.F90:31
31          DO i = 1, n
Aborted (core dumped)
...

In more detail:
...
Temporary breakpoint 1, MAIN__ () at test.F90:31
31          DO i = 1, n

Thread 1 "gdb" received signal SIGSEGV, Segmentation fault.
0x00000000006ab78e in evaluate_subexp_f (expect_type=0x0, exp=0x2067bf0,
pos=0x7fffffffd3bc, 
    noside=EVAL_NORMAL) at gdb/f-lang.c:938
938                           = TYPE_FIELD_ARTIFICIAL (value_type (arg1), tem -
1);
...

So tem is:
...
(gdb) p tem
$2 = 1
...
and value_type (arg1) is :
...
(gdb) p value_type (arg1)
$1 = (type *) 0x21cc140
(gdb) call recursive_dump_type ($1, 0)
type node 0x21cc140
name '<text gnu-indirect-function variable, no debug info>' (0xe5f160)
code 0x7 (TYPE_CODE_FUNC)
length 1
objfile 0x224c490
target_type 0x0
pointer_type 0x0
reference_type 0x0
type_chain 0x21cc140
instance_flags 0x0
flags
nfields 0 0x0
calling_convention 0
...

So, we do TYPE_FIELD_ARTIFICIAL($1, 0).

The macro TYPE_FIELD_ARTIFICIAL is defined like this:
...
#define TYPE_FIELD_ARTIFICIAL(thistype, n) FIELD_ARTIFICIAL((thistype)->field
(n))
...
so we try to access field 0 of type $1, but the type has no fields (see nfields
in the recursive type dump).

-- 
You are receiving this mail because:
You are on the CC list for the bug.