[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615
--- Comment #6 from Dominique d'Humieres ---
The following test does not need -fdefault-integer-8, but only -m32 to give a
wrong-code:
module mykinds
implicit none
integer, parameter :: ik1 = selected_int_kind (2)
integer, parameter :: ik2 = selected_int_kind (4)
! integer, parameter :: ik = selected_int_kind (8)
integer, parameter :: ik = selected_int_kind (12)
integer, parameter :: dp = selected_real_kind (15,300)
end module mykinds
module spec_xpr
use mykinds
implicit none
integer(ik2) c_size
contains
pure function tricky (str,ugly)
character(*), intent(in) :: str
integer(ik1) :: ia_ik1(len(str))
interface yoagly
pure function ugly(n)
use mykinds
implicit none
integer(ik), intent(in) :: n
complex(dp) :: ugly(3*n+2)
end function ugly
end interface yoagly
logical(ik) :: la(size (yoagly (size (ia_ik1, kind=ik
integer(ik) :: i
character(tricky_helper ((/(.TRUE._ik, i=1, size (la))/)) + c_size) ::
tricky
tricky = transfer (yoagly (1_ik), tricky)
end function tricky
pure function tricky_helper (lb)
logical(ik), intent(in) :: lb(:)
integer(ik) :: tricky_helper
tricky_helper = 2 * size (lb) + 3
end function tricky_helper
end module spec_xpr
module xtra_fun
implicit none
contains
pure function butt_ugly(n)
use mykinds
implicit none
integer(ik), intent(in) :: n
complex(dp) :: butt_ugly(3*n+2)
real(dp) pi, sq2
pi = 4 * atan (1.0_dp)
sq2 = sqrt (2.0_dp)
butt_ugly = cmplx (pi, sq2, dp)
end function butt_ugly
end module xtra_fun
program spec_test
use mykinds
use spec_xpr
use xtra_fun
implicit none
character(54) :: chr
c_size = 5
if (tricky ('Help me', butt_ugly) .ne. transfer (butt_ugly (1_ik), chr)) call
abort ()
end program spec_test
[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615
--- Comment #5 from Dominique d'Humieres ---
The following test (reduced from gfortran.dg/pr46588.f90)
function aufun(pm)
character(len = *) pm
character(len = *) aufun
print *, ">", pm, "<"
aufun = 'Oh' // trim(pm)
end function aufun
program ds
implicit none
character(len = 4) :: ins = ' no!'
character(len = 20) st, aufun
logical(4) :: tst
print *, ">", ins, "<"
st = aufun(ins)
print *, ">", st, "<"
end
gives
> no!<
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
at run time when compiled with -m32 -fdefault-integer-8.
I don't see any difference when comparing the outputs of -fdump-tree-original
with -m32 or -m32 -fdefault-integer-8, but I see the following difference in
the assembly files
--- pr46588_db.s2018-03-05 12:50:58.0 +0100
+++ pr46588_db_m32.s2018-03-05 12:50:41.0 +0100
@@ -3,8 +3,7 @@
# GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
# options passed: pr46588_db.f90 -fPIC -m32 -mmacosx-version-min=10.13.0
-# -mtune=corei7 -march=corei7 -fdefault-integer-8 -fdump-tree-original
-# -fverbose-asm
+# -mtune=corei7 -march=corei7 -fdump-tree-original -fverbose-asm
# -fintrinsic-modules-path
/opt/gcc/gcc8w/lib/gcc/x86_64-apple-darwin17.4.0/8.0.1/i386/finclude
# options enabled: -Wnonportable-cfstrings -fPIC
# -faggressive-loop-optimizations -fasynchronous-unwind-tables
@@ -255,16 +254,14 @@ L2$pb:
call__gfortran_st_write_done#
addl$16, %esp #,
# pr46588_db.f90:14:st = aufun(ins)
- subl$12, %esp #,
pushl $4 #
leal_ins.3702-L2$pb(%ebx), %eax #, tmp97
pushl %eax# tmp97
- pushl $0 #
pushl $20 #
leal-48(%ebp), %eax #, tmp98
pushl %eax# tmp98
call_aufun_ #
- addl$32, %esp #,
+ addl$16, %esp #,
movl-48(%ebp), %eax # MEM[(c_char * {ref-all})], tmp99
movl%eax, -28(%ebp) # tmp99, MEM[(c_char * {ref-all})]
movl-44(%ebp), %eax # MEM[(c_char * {ref-all})], tmp100
[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615 --- Comment #4 from Thomas Koenig --- On a big-endian system (gcc110), there are segfaults. Looking into char_reshape_1.f90 yields Breakpoint 1, _gfortran_reshape_char (ret=0xfffef130, ret_length=4294897968, source=0x0, shape=0x9, pad=0xfffef1d8, order=0xfffef1b4, source_length=4294898064, pad_length=4294898028) at ../../../../trunk/libgfortran/intrinsics/reshape_generic.c:381 Looking at the tree dump gives something very strange: struct array01_unknown * D.2133; [...] D.2133 = typedef integer(kind=4) [3]; and then _gfortran_reshape_char (, 9, D.2133, D.2145, D.2147, D.2153, 9, 9); (wrong prototype, gdb is apparently showing the function in the tail call). So it seems that the stirng length constant simply has the wrong length, it occupies eight bytes where it should occupy four. Janne?
[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615 --- Comment #3 from Thomas Koenig --- This has to be some issue with the charlen - possibly a default integer being used in the wrong place after r256322. I don't usually build 32-bit-capable compilers on my home system (too slow), let's see if I can get something built on the compile farm.
[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615 Dominique d'Humieres changed: What|Removed |Added Blocks||32770 --- Comment #2 from Dominique d'Humieres --- s/comment 1/comment 0/ Referenced Bugs: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=32770 [Bug 32770] [Meta-bug] -fdefault-integer-8 issues
[Bug fortran/84615] [8 Regression] Executable Segfault for tests compiled with -fdefault-integer-8 and -m32
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84615 Dominique d'Humieres changed: What|Removed |Added Priority|P3 |P4 Status|UNCONFIRMED |NEW Known to work||7.3.0 Keywords||wrong-code Last reconfirmed||2018-02-28 CC||jb at gcc dot gnu.org, ||tkoenig at gcc dot gnu.org Ever confirmed|0 |1 Target Milestone|--- |8.0 Known to fail||8.0 --- Comment #1 from Dominique d'Humieres --- Note that the list given in comment 1 may be non-exhaustive and/or contain a few errors.
