[Bug ada/60174] ICE on ACATS cc3305a

[Bug ada/60174] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Can you check from the -fdump-tree-all dumps where i_34(ab) and i_399(ab)
start to have overlapping life-ranges?

[Bug ada/60174] ICE on ACATS cc3305a

[mikpelinux at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

Mikael Pettersson <mikpelinux at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mikpelinux at gmail dot com

--- Comment #2 from Mikael Pettersson <mikpelinux at gmail dot com> ---
I got the exact same error on armv5tel-linux-gnueabi with 4.9-20140209 aka
r207646.  The previous weekly snapshot, 4.9-20140202 aka r207409, was Ok.

[Bug ada/60174] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at gcc dot gnu.org

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Looks like so.  Can you attach -fdump-tree-dom1-details?

[Bug ada/60174] ICE on ACATS cc3305a

[bernd.edlinger at hotmail dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #6 from Bernd Edlinger <bernd.edlinger at hotmail dot de> ---
Created attachment 32123
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32123&action=edit
output from -fdump-tree-dom1-details

(In reply to Richard Biener from comment #5)
> Looks like so.  Can you attach -fdump-tree-dom1-details?

sure, here you are.

[Bug ada/60174] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Ok, I believe this is caused by some jump-threading threading over a copy of
an SSA name used in abnormal context.  You can try verifying if
-fno-tree-dominator-opts fixes this (maybe you also need -fno-tree-vrp, not
sure)

[Bug ada/60174] ICE on ACATS cc3305a

[bernd.edlinger at hotmail dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #8 from Bernd Edlinger <bernd.edlinger at hotmail dot de> ---
(In reply to Richard Biener from comment #7)
> Ok, I believe this is caused by some jump-threading threading over a copy of
> an SSA name used in abnormal context.  You can try verifying if
> -fno-tree-dominator-opts fixes this (maybe you also need -fno-tree-vrp, not
> sure)

Yes, if I add -fno-tree-dominator-opts (_before_ -largs) to the gnatmake
command line, the ICE is gone, and the test case PASSES.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-02-13
          Component|ada                         |tree-optimization
                 CC|                            |ebotcazou at gcc dot gnu.org
           Assignee|law at redhat dot com              |unassigned at gcc dot gnu.org
     Ever confirmed|0                           |1
            Summary|ICE on ACATS cc3305a        |[4.9 regression] ICE on
                   |                            |ACATS cc3305a
   Target Milestone|---                         |4.9.0

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Probably a missing SSA_NAME_OCCURS_IN_ABNORMAL_PHI check somewhere.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[law at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #10 from Jeffrey A. Law <law at redhat dot com> ---
That was my thought as well.  I've got some unexpected family stuff to deal
with over the next couple days, but I don't see any reason why this can't be
nailed down early next week.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #11 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Reproducible on x86{-64}/Linux with the following procedure:

 - copy $srcdir/gcc/testsuite/ada/acats/support/repbody.ada to $buildir
 - gnatchop repbody.ada
 - copy $srcdir/gcc/testsuite/ada/acats/support/repspec.ada to $buildir
 - gnatchop repspec.ada
 - copy $builddir/gcc/ada/rts/system.ads to $buildir
 - edit system.ads and change ZCX_By_Default to False
 - compile

eric@polaris:~/build/gcc/native> gcc/xgcc -Bgcc -S cc3305a.adb -O2

Unable to coalesce ssa_names 34 and 399 which are marked as MUST COALESCE.
i_34(ab) and  i_399(ab)
+===========================GNAT BUG DETECTED==============================+
| 4.9.0 20140213 (experimental) [trunk revision 207749] (x86_64-suse-linux) GCC
error:|
| SSA corruption                                                           |
| Error detected around cc3305a.adb:35:1                                   |
| Please submit a bug report; see http://gcc.gnu.org/bugs.html.            |
| Use a subject line meaningful to you and us to track the bug.            |
| Include the entire contents of this bug box in the report.               |
| Include the exact gcc or gnatmake command that you entered.              |
| Also include sources listed below in gnatchop format                     |
| (concatenated together with no headers between files).                   |
+==========================================================================+

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Eric Botcazou from comment #11)
> Reproducible on x86{-64}/Linux with the following procedure:
> 
>  - copy $srcdir/gcc/testsuite/ada/acats/support/repbody.ada to $buildir
>  - gnatchop repbody.ada
>  - copy $srcdir/gcc/testsuite/ada/acats/support/repspec.ada to $buildir
>  - gnatchop repspec.ada
>  - copy $builddir/gcc/ada/rts/system.ads to $buildir
>  - edit system.ads and change ZCX_By_Default to False
>  - compile

Unfortunately that's always telling me

error: "gnat1drv.adb" must be recompiled ("system.ads" has been modified)
error: "s-stalib.adb" must be recompiled ("system.ads" has been modified)
error: "a-except.adb" must be recompiled ("system.ads" has been modified)
...

(recurring problem when I try to re-build gnat1 or all-gcc in my
dev-tree after svn up)

I wonder if this is a missing dependency.

> eric@polaris:~/build/gcc/native> gcc/xgcc -Bgcc -S cc3305a.adb -O2
> 
> Unable to coalesce ssa_names 34 and 399 which are marked as MUST COALESCE.
> i_34(ab) and  i_399(ab)
> +===========================GNAT BUG DETECTED==============================+
> | 4.9.0 20140213 (experimental) [trunk revision 207749] (x86_64-suse-linux)
> GCC error:|
> | SSA corruption                                                           |
> | Error detected around cc3305a.adb:35:1                                   |
> | Please submit a bug report; see http://gcc.gnu.org/bugs.html.            |
> | Use a subject line meaningful to you and us to track the bug.            |
> | Include the entire contents of this bug box in the report.               |
> | Include the exact gcc or gnatmake command that you entered.              |
> | Also include sources listed below in gnatchop format                     |
> | (concatenated together with no headers between files).                   |
> +==========================================================================+

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[bernd.edlinger at hotmail dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #13 from Bernd Edlinger <bernd.edlinger at hotmail dot de> ---
you can also build a cross-gcc/gnat for this spec.
all you need for it, are the system root files, but I can give them to you.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #14 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Unfortunately that's always telling me
> 
> error: "gnat1drv.adb" must be recompiled ("system.ads" has been modified)
> error: "s-stalib.adb" must be recompiled ("system.ads" has been modified)
> error: "a-except.adb" must be recompiled ("system.ads" has been modified)

I don't see how this can happen if you strictly follow the guidelines (when I
write "compile", I mean issue the command line just below).

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ebotcazou at gcc dot gnu.org

--- Comment #15 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
I'll have a look myself.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #32122|0                           |1
        is obsolete|                            |
  Attachment #32123|0                           |1
        is obsolete|                            |

--- Comment #16 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Created attachment 32161
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32161&action=edit
Reduced testcase

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #17 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Eric Botcazou from comment #14)
> > Unfortunately that's always telling me
> > 
> > error: "gnat1drv.adb" must be recompiled ("system.ads" has been modified)
> > error: "s-stalib.adb" must be recompiled ("system.ads" has been modified)
> > error: "a-except.adb" must be recompiled ("system.ads" has been modified)
> 
> I don't see how this can happen if you strictly follow the guidelines (when
> I write "compile", I mean issue the command line just below).

Then I cannot reproduce it, but I have to alter the commandline as

/abuild/rguenther/trunk-g> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2 
fatal error, run-time library not installed correctly
cannot locate file system.ads
compilation abandoned

thus I add -I gcc/

> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
cc3305a.adb:74:11: run-time library configuration error
cc3305a.adb:74:11: file s-soflin.ads not found
cc3305a.adb:74:11: entity "System.Soft_Links.Abort_Undefer" not available
compilation abandoned

at which point I need to add -I gcc/ada/rts but then the testcase works.

/abuild/rguenther/trunk-g> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/ -I
gcc/ada/rts
/abuild/rguenther/trunk-g>

Re: [Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[Arnaud Charlet]

> > ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> cc3305a.adb:74:11: run-time library configuration error
> cc3305a.adb:74:11: file s-soflin.ads not found
> cc3305a.adb:74:11: entity "System.Soft_Links.Abort_Undefer" not
> available
> compilation abandoned
> 
> at which point I need to add -I gcc/ada/rts but then the testcase works.
> 
> /abuild/rguenther/trunk-g> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> -I
> gcc/ada/rts
> /abuild/rguenther/trunk-g>

You need to use -I . -I gcc/ada/rts (and no -I gcc/)

Arno

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[charlet at adacore dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #18 from charlet at adacore dot com <charlet at adacore dot com> ---
> > ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> cc3305a.adb:74:11: run-time library configuration error
> cc3305a.adb:74:11: file s-soflin.ads not found
> cc3305a.adb:74:11: entity "System.Soft_Links.Abort_Undefer" not
> available
> compilation abandoned
> 
> at which point I need to add -I gcc/ada/rts but then the testcase works.
> 
> /abuild/rguenther/trunk-g> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> -I
> gcc/ada/rts
> /abuild/rguenther/trunk-g>

You need to use -I . -I gcc/ada/rts (and no -I gcc/)

Arno

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #19 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to charlet@adacore.com from comment #18)
> > > ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> > cc3305a.adb:74:11: run-time library configuration error
> > cc3305a.adb:74:11: file s-soflin.ads not found
> > cc3305a.adb:74:11: entity "System.Soft_Links.Abort_Undefer" not
> > available
> > compilation abandoned
> > 
> > at which point I need to add -I gcc/ada/rts but then the testcase works.
> > 
> > /abuild/rguenther/trunk-g> ./gcc/xgcc -Bgcc -S cc3305a.adb -O2  -I gcc/
> > -I
> > gcc/ada/rts
> > /abuild/rguenther/trunk-g>
> 
> You need to use -I . -I gcc/ada/rts (and no -I gcc/)
> 
> Arno

Same result (system.ads and rep*.adb were in gcc/, but even moving them
to $builddir and using -I. doesn't work).  Same result == testcase passes.

Maybe I have to rebuild gnat1 using that modified system.ads?  Even
when I change the master copy in gcc/ada/rts/ the error does not reproduce
for me.  But I have not re-built gnat1 for quite some time beacuse of the
errors that I should recompile stuff.

Re: [Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[Arnaud Charlet]

> Same result (system.ads and rep*.adb were in gcc/, but even moving them
> to $builddir and using -I. doesn't work).  Same result == testcase passes.
> 
> Maybe I have to rebuild gnat1 using that modified system.ads?  Even

No, you don't need to do that, gnat1 reads system.ads at run time.

Assuming you have done a make install, then under the directory
where you've done the gnatchop and copy of system.ads, it should really
be a trivial matter of:

gcc -c <test case.adb>

Nothing more nothing less. (You can then of course debug using gnat1
once you can reproduce with the above).

Arno

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[charlet at adacore dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #20 from charlet at adacore dot com <charlet at adacore dot com> ---
> Same result (system.ads and rep*.adb were in gcc/, but even moving them
> to $builddir and using -I. doesn't work).  Same result == testcase passes.
> 
> Maybe I have to rebuild gnat1 using that modified system.ads?  Even

No, you don't need to do that, gnat1 reads system.ads at run time.

Assuming you have done a make install, then under the directory
where you've done the gnatchop and copy of system.ads, it should really
be a trivial matter of:

gcc -c <test case.adb>

Nothing more nothing less. (You can then of course debug using gnat1
once you can reproduce with the above).

Arno

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #21 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
It's reassociation, namely optimize_range_tests, doing SSA_NAME propagation
when building ranges in init_range_entry.  The immediate fix is:

Index: tree-ssa-reassoc.c
===================================================================
--- tree-ssa-reassoc.c  (revision 207796)
+++ tree-ssa-reassoc.c  (working copy)
@@ -1840,7 +1840,8 @@ init_range_entry (struct range_entry *r,

       if (exp != NULL_TREE)
        {
-         if (TREE_CODE (exp) != SSA_NAME)
+         if (TREE_CODE (exp) != SSA_NAME
+             || SSA_NAME_OCCURS_IN_ABNORMAL_PHI (exp))
            break;

          stmt = SSA_NAME_DEF_STMT (exp);

The range test is still optimized, but the SSA_NAME is not replaced anymore.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #22 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Patch preapproved.

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

--- Comment #23 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Author: ebotcazou
Date: Tue Feb 18 23:15:27 2014
New Revision: 207859

URL: http://gcc.gnu.org/viewcvs?rev=207859&root=gcc&view=rev
Log:
    PR tree-optimization/60174
    * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
    statement of an SSA_NAME that occurs in an abnormal PHI node.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-ssa-reassoc.c

[Bug tree-optimization/60174] [4.9 regression] ICE on ACATS cc3305a

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60174

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
                 CC|ebotcazou at gcc dot gnu.org       |
         Resolution|---                         |FIXED

--- Comment #24 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
Thanks for reporting the problem.