[Bug middle-end/54041] New: -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[Bug middle-end/54041] New: -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[ksop at hot dot ee]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

             Bug #: 54041
           Summary: -mshort ICE in convert_memory_address_addr_space, at
                    explow.c:327
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ksop@hot.ee


Appears when pointer+index is evaluated, with sizetype being short.

Also tried with one convert_memory_address_addr_space() call commented out, in
which case no ICE is produced. 4.6.3, 4.5.4 affected as well. An -mshort target
makes an interesting testing tool.

$ m68k-linux-gnu-gcc-4.7.1 -v
Using built-in specs.
COLLECT_GCC=m68k-linux-gnu-gcc-4.7.1
COLLECT_LTO_WRAPPER=/opt/gnu/libexec/gcc/m68k-linux-gnu/4.7.1/lto-wrapper
Target: m68k-linux-gnu
Configured with: ./configure --target=m68k-linux-gnu --build=x86_64-linux-gnu
--enable-languages=c --prefix=/opt/gnu
Thread model: posix
gcc version 4.7.1 (GCC) 

$ m68k-linux-gnu-gcc-4.7.1 -mshort -O -S test.i
test.i: In function 'fn':
test.i:5:2: internal compiler error: in convert_memory_address_addr_space, at
explow.c:327
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[ksop at hot dot ee]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #1 from Kalev Soikonen <ksop at hot dot ee> 2012-07-20 01:59:05 UTC ---
Created attachment 27840
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27840
testcase

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[ksop at hot dot ee]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #2 from Kalev Soikonen <ksop at hot dot ee> 2012-07-20 02:01:21 UTC ---
Created attachment 27841
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=27841
melts the ICE

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Mikael Pettersson <mikpe at it dot uu.se> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpe at it dot uu.se

--- Comment #3 from Mikael Pettersson <mikpe at it dot uu.se> 2012-07-20 10:22:03 UTC ---
Was this originally seen with an out-of-tree port?

m68k-linux-gcc -mshort is an unusual combination, and I would expect it to be
essentially unsupported as -mshort changes the ABI.  For m68k-elf the option
might make sense, but I wasn't able to reproduce the ICE for m68k-elf, only
m68k-linux.

The ICE comes from explow.c:

rtx
convert_memory_address_addr_space (enum machine_mode to_mode ATTRIBUTE_UNUSED,
                                   rtx x, addr_space_t as ATTRIBUTE_UNUSED)
{
#ifndef POINTERS_EXTEND_UNSIGNED
  gcc_assert (GET_MODE (x) == to_mode || GET_MODE (x) == VOIDmode);
  return x;
#else /* defined(POINTERS_EXTEND_UNSIGNED) */

m68k doesn't define P_E_U, but with -mshort to_mode apparently ends up being
shorter, so the above assertion fails.

The documentation for P_E_U states:

"
You need not define this macro if the @code{ptr_mode}, @code{Pmode}
and @code{word_mode} are all the same width.
"

-mshort changes word_mode, invalidating the condition for not defining P_E_U.

In this case a truncation is asked for.  convert_memory_address_addr_space does
perform truncations, but only when P_E_U is defined.  Defining P_E_U to 0
around c_m_a_a_s eliminated the ICE and generated reasonable-looking code.

I'm not sure if the bug is that the m68k backend doesn't define P_E_U even
though is permits -mshort, or that c_m_a_a_s is wrong and should perform
truncations also when P_E_U is undefined.

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #4 from Andreas Schwab <schwab@linux-m68k.org> 2012-07-20 10:54:41 UTC ---
-mshort doesn't change word_mode, UNITS_PER_WORD is fixed at 4.  -mshort only
changes sizeof(int).

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #5 from Mikael Pettersson <mikpe at it dot uu.se> 2012-07-20 11:33:14 UTC ---
Ok, in that case I'd say the m68k backend is entitled to not define P_E_U, and
it's really c_m_a_a_s (or something related to it) that's broken.

[Bug middle-end/54041] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #6 from Andreas Schwab <schwab@linux-m68k.org> 2012-07-20 12:37:41 UTC ---
expr.c:7576:      tmp = convert_memory_address_addr_space (tmode, tmp, as);

This looks bogus, tmp (which is the expanded offset) isn't a memory address.

[Bug middle-end/54041] [4.7/4.8 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-07-20
      Known to work|                            |4.6.4
            Summary|-mshort ICE in              |[4.7/4.8 regression]
                   |convert_memory_address_addr |-mshort ICE in
                   |_space, at explow.c:327     |convert_memory_address_addr
                   |                            |_space, at explow.c:327
     Ever Confirmed|0                           |1
      Known to fail|                            |4.7.1

--- Comment #7 from Andreas Schwab <schwab@linux-m68k.org> 2012-07-20 13:15:43 UTC ---
Before 4.7 the array reference was expanded to (int *) &r + (unsigned int)
((unsigned int) i * 2), now it is preserved as &r[i], causing the ADDR_EXPR to
contain ARRAY_REF of (r,i) where it previously contained r only.

[Bug middle-end/54041] [4.7/4.8 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.2

[Bug middle-end/54041] [4.7/4.8 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |m68k-linux-gnu
           Priority|P3                          |P4

[Bug middle-end/54041] [4.7/4.8 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.7.2                       |4.7.3

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-09-20 10:19:39 UTC ---
GCC 4.7.2 has been released.

[Bug middle-end/54041] [4.7/4.8/4.9 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.7.3                       |4.7.4

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> 2013-04-11 07:59:26 UTC ---
GCC 4.7.3 is being released, adjusting target milestone.

[Bug middle-end/54041] [4.7/4.8/4.9 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P4                          |P5
                 CC|                            |law at redhat dot com
           Assignee|unassigned at gcc dot gnu.org      |law at redhat dot com

--- Comment #10 from Jeffrey A. Law <law at redhat dot com> ---
I believe the problem is expand_expr_addr_expr_1 assumes that expand_expr will
produce a result in the requested mode.  Unfortunately, expand_expr has always
been free to return an object with a different mode.

If we use convert_modes in the obvious way when tmode != GET_MODE (tmp) we no
longer get the ICE and the resulting code looks reasonably correct.

Of course we don't really have a good way to test -mshort.  I'm not even sure
if it's used anymore in real world scenarios.

I'll do a bootstrap and test on x86_64, but I obviously realize that's not
going to give this change any real coverage testing.

[Bug middle-end/54041] [4.7/4.8/4.9 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[law at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

--- Comment #11 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Author: law
Date: Tue Feb 11 13:58:46 2014
New Revision: 207689

URL: http://gcc.gnu.org/viewcvs?rev=207689&root=gcc&view=rev
Log:
    PR middle-end/54041
    * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
    object with an undesirable mode.

    PR middle-end/54041
    * gcc.target/m68k/pr54041.c: New test.

Added:
    trunk/gcc/testsuite/gcc.target/m68k/pr54041.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/expr.c
    trunk/gcc/testsuite/ChangeLog

[Bug middle-end/54041] [4.7/4.8/4.9 regression] -mshort ICE in convert_memory_address_addr_space, at explow.c:327

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54041

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #12 from Jeffrey A. Law <law at redhat dot com> ---
Fixed on trunk.  No plans to backport fix to release branches.