public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug fortran/108872] New: ICE in main_block_label, missing label when assigning a derived type whose component has a defined assignment
@ 2023-02-21 15:11 trnka at scm dot com
2023-02-21 18:18 ` [Bug fortran/108872] " anlauf at gcc dot gnu.org
0 siblings, 1 reply; 2+ messages in thread
From: trnka at scm dot com @ 2023-02-21 15:11 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108872
Bug ID: 108872
Summary: ICE in main_block_label, missing label when assigning
a derived type whose component has a defined
assignment
Product: gcc
Version: 12.2.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: trnka at scm dot com
Target Milestone: ---
gfortran (tested versions 9 through 12.2.1 20221121) crashes when compiling the
following testcase because the "bb" pointer is null in main_block_label()
called from cleanup_dead_labels() in tree-cfg.c:
module TestModule
implicit none
private
type :: A
integer :: i = 0
contains
procedure :: NewCopyOther
generic :: assignment(=) => NewCopyOther
end type
type :: B
type(A) :: aa
end type
type :: C
type(B) :: x
contains
procedure :: Run
end type
contains
subroutine NewCopyOther(self, other)
class(A), intent(out) :: self
class(A), intent(in) :: other
self%i = other%i
end subroutine
subroutine Run(self)
class(C), intent(inout) :: self
type(B) :: y
goto 2
2 y = self%x
end subroutine
end module
Apparently, the Fortran frontend never emits the "2" label as long as it's
placed on a line containing an assignment of a derived type (B) which has a
component with a defined assignment operator (A). The "original" dump for run()
looks like the following:
__attribute__((fn spec (". w ")))
void run (struct __class_testmodule_C_t & restrict self)
{
struct b y;
{
struct b b.2;
b.2.aa.i = 0;
y = b.2;
}
goto __label_000002;
y = self->_data->x;
{
struct __class_testmodule_A_t class.0;
struct __class_testmodule_A_t class.1;
class.0._vptr = (struct __vtype_testmodule_A * {ref-all})
&__vtab_testmodule_A;
class.0._data = &y.aa;
class.1._vptr = (struct __vtype_testmodule_A * {ref-all})
&__vtab_testmodule_A;
class.1._data = &self->_data->x.aa;
newcopyother (&class.0, &class.1);
}
}
__label_000002 is thus used in a goto even though it's nowhere to be seen.
I guess bug 95613 might be related but probably not a duplicate as that one has
nothing to do with defined assignments.
^ permalink raw reply [flat|nested] 2+ messages in thread
* [Bug fortran/108872] ICE in main_block_label, missing label when assigning a derived type whose component has a defined assignment
2023-02-21 15:11 [Bug fortran/108872] New: ICE in main_block_label, missing label when assigning a derived type whose component has a defined assignment trnka at scm dot com
@ 2023-02-21 18:18 ` anlauf at gcc dot gnu.org
0 siblings, 0 replies; 2+ messages in thread
From: anlauf at gcc dot gnu.org @ 2023-02-21 18:18 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108872
anlauf at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Keywords| |ice-on-valid-code
Last reconfirmed| |2023-02-21
Ever confirmed|0 |1
--- Comment #1 from anlauf at gcc dot gnu.org ---
Confirmed. All gcc versions tested (>=7) fail.
Note that pr95613 is about invalid code, while the testcase in comment#0
appears to be valid Fortran.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-02-21 18:18 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-21 15:11 [Bug fortran/108872] New: ICE in main_block_label, missing label when assigning a derived type whose component has a defined assignment trnka at scm dot com
2023-02-21 18:18 ` [Bug fortran/108872] " anlauf at gcc dot gnu.org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).