[Bug middle-end/51442] New: volatile bitfields broken on arm

[Bug middle-end/51442] New: volatile bitfields broken on arm

[bconoboy at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

             Bug #: 51442
           Summary: volatile bitfields broken on arm
    Classification: Unclassified
           Product: gcc
           Version: 4.6.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: bconoboy@gmail.com

[Bug middle-end/51442] volatile bitfields broken on arm

[bconoboy at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

--- Comment #1 from Brendan Conoboy <bconoboy at gmail dot com> 2011-12-06 21:02:22 UTC ---
This bug is in reference to the patch that Diego Novillo accepted into 4.7
about 10 months ago:

http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01477.html

The issue is still present in gcc 4.6.x and the same patch resolves the issue. 
This is a problem for all gcc 4.6-based ARM targets.

In Debian, the discussion and resolution is here:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=625825

Linaro and Ubuntu use the same patch.  Discussion here:

https://bugs.launchpad.net/gcc-linaro/+bug/675347

We've been carrying the patch in Fedora-ARM as well.  Without the patch applied
the USB stack is broken on ARM, as are various KDE/Qt builds.

[Bug middle-end/51442] volatile bitfields broken on arm

[bconoboy at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

--- Comment #2 from Brendan Conoboy <bconoboy at gmail dot com> 2011-12-06 21:12:46 UTC ---
Created attachment 26010
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26010
Only use BLKmode for volatile accesses which are not naturally aligned.

Per Julian Brown's original email at
http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01477.html:

    gcc/
    * expr.c (expand_expr_real_1): Only use BLKmode for volatile
    accesses which are not naturally aligned.

[Bug middle-end/51442] volatile bitfields broken on arm

[rearnsha at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

--- Comment #3 from Richard Earnshaw <rearnsha at gcc dot gnu.org> 2011-12-07 10:55:17 UTC ---
(In reply to comment #2)
> Created attachment 26010 [details]
> Only use BLKmode for volatile accesses which are not naturally aligned.
> 
> Per Julian Brown's original email at
> http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01477.html:
> 
>     gcc/
>     * expr.c (expand_expr_real_1): Only use BLKmode for volatile
>     accesses which are not naturally aligned.

If this patch is backported then so must:

http://gcc.gnu.org/ml/gcc-patches/2011-11/msg01390.html

[Bug middle-end/51442] volatile bitfields broken on arm

[bconoboy at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

--- Comment #4 from Brendan Conoboy <bconoboy at gmail dot com> 2011-12-10 01:49:25 UTC ---
FYI, I've checked out svn://gcc.gnu.org/svn/gcc/branches/gcc-4_6-branch
revision 182172, run and rerun the complete testsuite before and after this
patch (As well as the patch Richard pointed at:
http://gcc.gnu.org/ml/gcc-patches/2011-11/msg01390/volatile-bitfields-1110.patch)
on an x86_64 host.  The results are as follows:

objc, gfortran, libstdc++-v3, libjava, libmudflap, libffi, libgomp: identical
gcc: 2 new passes (These are from the testcase added by the second patch). 
Otherwise identical.

To me this suggests there is very little risk in adding these patches to the
branch.  Likewise, there is a great deal to be gained for ARM users.

[Bug middle-end/51442] volatile bitfields broken on arm

[ramana at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51442

Ramana Radhakrishnan <ramana at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |ramana at gcc dot gnu.org
         Resolution|                            |FIXED
   Target Milestone|---                         |4.6.3

--- Comment #5 from Ramana Radhakrishnan <ramana at gcc dot gnu.org> 2012-01-11 12:12:07 UTC ---
This should now be fixed by the commit as indicated below - 

Brendan - can you confirm this is the case ? 



Author: dj
Date: Thu Dec 22 19:36:46 2011
New Revision: 182635

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=182635
Log:
2011-12-22  Doug Kwan  <dougkwan@google.com>

    Backport from mainline
    2011-03-23  Julian Brown  <julian@codesourcery.com>

    * expr.c (expand_expr_real_1): Only use BLKmode for volatile
    accesses which are not naturally aligned.

    2011-11-20  Joey Ye  <joey.ye@arm.com>

    * expr.c (expand_expr_real_1): Correctly handle strict volatile
    bitfield loads smaller than mode size.

2011-12-22  Doug Kwan  <dougkwan@google.com>

    Backport from mainline
    2011-11-20  Joey Ye  <joey.ye@arm.com>

    * gcc.dg/volatile-bitfields-1.c: New.

Added:
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/volatile-bitfields-1.c
Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/expr.c
    branches/gcc-4_6-branch/gcc/testsuite/ChangeLog