[Bug gcov-profile/51297] New: [4.7 regressions] Many gcov tests FAIL on Tru64 UNIX, Solaris 8

[Bug gcov-profile/51297] New: [4.7 regressions] Many gcov tests FAIL on Tru64 UNIX, Solaris 8

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

             Bug #: 51297
           Summary: [4.7 regressions] Many gcov tests FAIL on Tru64 UNIX,
                    Solaris 8
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: gcov-profile
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ro@gcc.gnu.org
                CC: nathan@gcc.gnu.org
              Host: alpha-dec-osf5.1b, *-*-solaris2.8
            Target: alpha-dec-osf5.1b, *-*-solaris2.8
             Build: alpha-dec-osf5.1b, *-*-solaris2.8


Between 20111110 (r181259) and 20111114 (r181350), many gcov tests started to
FAIL on Solaris 8 (SPARC and x86) and Tru64 UNIX V5.1B:

FAIL: g++.dg/gcov/gcov-1.C gcov failed: 
FAIL: g++.dg/gcov/gcov-1.C gcov failed: 
FAIL: g++.dg/gcov/gcov-10.C gcov failed: 
FAIL: g++.dg/gcov/gcov-10.C gcov failed: 
FAIL: g++.dg/gcov/gcov-11.C gcov failed: 
FAIL: g++.dg/gcov/gcov-11.C gcov failed: 
FAIL: g++.dg/gcov/gcov-2.C gcov failed: 
FAIL: g++.dg/gcov/gcov-2.C gcov failed: 
FAIL: g++.dg/gcov/gcov-3.C gcov failed: 
FAIL: g++.dg/gcov/gcov-3.C gcov failed: 
FAIL: g++.dg/gcov/gcov-4.C gcov failed: 
FAIL: g++.dg/gcov/gcov-4.C gcov failed: 
FAIL: g++.dg/gcov/gcov-5.C gcov failed: 
FAIL: g++.dg/gcov/gcov-5.C gcov failed: 
FAIL: g++.dg/gcov/gcov-7.C gcov failed: 
FAIL: g++.dg/gcov/gcov-7.C gcov failed: 
FAIL: gcc.misc-tests/gcov-1.c gcov failed: 
FAIL: gcc.misc-tests/gcov-10.c gcov failed: 
FAIL: gcc.misc-tests/gcov-10b.c gcov failed: 
FAIL: gcc.misc-tests/gcov-11.c gcov failed: 
FAIL: gcc.misc-tests/gcov-12.c gcov failed: 
FAIL: gcc.misc-tests/gcov-13.c gcov failed: 
FAIL: gcc.misc-tests/gcovpart-13b.c gcov failed: 
FAIL: gcc.misc-tests/gcov-14.c (test for excess errors)
WARNING: gcc.misc-tests/gcov-14.c compilation failed to produce executable
FAIL: gcc.misc-tests/gcov-14.c gcov failed: 
FAIL: gcc.misc-tests/gcov-15.c gcov failed: 
FAIL: gcc.misc-tests/gcov-2.c gcov failed: 
FAIL: gcc.misc-tests/gcov-3.c gcov failed: 
FAIL: gcc.misc-tests/gcov-4.c gcov failed: 
FAIL: gcc.misc-tests/gcov-4b.c gcov failed: 
FAIL: gcc.misc-tests/gcov-5b.c gcov failed: 
FAIL: gcc.misc-tests/gcov-6.c gcov failed: 
FAIL: gcc.misc-tests/gcov-7.c gcov failed: 
FAIL: gcc.misc-tests/gcov-8.c gcov failed: 
FAIL: gcc.misc-tests/gcov-9.c gcov failed: 

The message is pretty useless since it states no reason.

The problem is that gcov dies with a SEGV:

> /var/gcc/regression/trunk/8-gcc-gas/build/gcc/gcov gcov-1.c
Segmentation Fault

Program received signal SIGSEGV, Segmentation fault.
0x08056d17 in name_search (a_=0x80a2f70, b_=0x7ffffff8)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:838
838       return strcmp (a, b->name);
(gdb) where
#0  0x08056d17 in name_search (a_=0x80a2f70, b_=0x7ffffff8)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:838
#1  0xbf6fc6ae in bsearch () from /usr/lib/libc.so.1
#2  0x08056d88 in find_source (file_name=0x80a2f70 <error reading variable>)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:866
#3  0x0808066e in read_graph_file (argc=2, argv=0x80479f4)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:1017
#4  process_file (argc=2, argv=0x80479f4)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:571
#5  main (argc=2, argv=0x80479f4) at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:423

The second argument to name_search is invalid, it seems.

  Rainer

[Bug gcov-profile/51297] [4.7 regressions] Many gcov tests FAIL on Tru64 UNIX, Solaris 8

[nathan at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #1 from Nathan Sidwell <nathan at gcc dot gnu.org> 2011-11-24 18:30:40 UTC ---
The line numbers in the backtrace don't seem to correspond to current sources. 
for instance line 866 is the definition of find_source, not the location of one
of the two bsearch calls.

which of the two bsearch calls is blowing up?

this one:
 name_map = (name_map_t *)bsearch
    (file_name, names, n_names, sizeof (*names), name_search);

or this one:
  name_map = (name_map_t *)bsearch
    (canon, names, n_names, sizeof (*names), name_search);

What are the values being passed to the bsearch call?

Oh, I see that it appears the string being passed to 'find_source' is
unreadable:
#2  0x08056d88 in find_source (file_name=0x80a2f70 <error reading variable>)
    at /vol/gcc/src/hg/trunk/local/gcc/gcov.c:866

What does gcov-dump -lo tell you about the gcno file?

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|alpha-dec-osf5.1b,          |alpha-dec-osf5.1b,
                   |*-*-solaris2.8              |*-*-solaris2.[89]
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-11-24
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org
               Host|alpha-dec-osf5.1b,          |alpha-dec-osf5.1b,
                   |*-*-solaris2.8              |*-*-solaris2.[89]
   Target Milestone|---                         |4.7.0
            Summary|[4.7 regressions] Many gcov |[4.7 regression] Many gcov
                   |tests FAIL on Tru64 UNIX,   |tests FAIL on Tru64,
                   |Solaris 8                   |Solaris 8 and 9
     Ever Confirmed|0                           |1
              Build|alpha-dec-osf5.1b,          |alpha-dec-osf5.1b,
                   |*-*-solaris2.8              |*-*-solaris2.[89]

--- Comment #2 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-11-24 18:36:36 UTC ---
I have them on SPARC/Solaris 8 and 9.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-11-24 18:40:43 UTC ---
> What are the values being passed to the bsearch call?

Very likely the problem indeed.  qsort is known to be very picky on Solaris 8
and 9, in the sense that the comparer function must impose a total order on the
array or else the function crashes.  I presume it's the same for bsearch.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[nathan at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #4 from Nathan Sidwell <nathan at gcc dot gnu.org> 2011-11-24 19:08:24 UTC ---
the names being entered into the array are unique, so there is a total ordering
-- I think that's a red herring.  I think the problem is the string read from
the data file is corrupted in some way.  Is that clueful enough?

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-11-24 19:46:20 UTC ---
> the names being entered into the array are unique, so there is a total ordering
> -- I think that's a red herring.  I think the problem is the string read from
> the data file is corrupted in some way.  Is that clueful enough?

Very likely a debugging artifact.  Here's my backtrace:

#0  name_search (a_=0x554a0, b_=0x0)
    at /nile.build/botcazou/gcc-head/src/gcc/gcov.c:838
#1  0xff2360e0 in bsearch () from /usr/lib/libc.so.1
#2  0x00015f34 in find_source (file_name=0x554a0 "gcov-1.c")
    at /nile.build/botcazou/gcc-head/src/gcc/gcov.c:866
#3  0x0003a828 in read_graph_file ()
    at /nile.build/botcazou/gcc-head/src/gcc/gcov.c:1017
#4  process_file (file_name=<optimized out>)
    at /nile.build/botcazou/gcc-head/src/gcc/gcov.c:571
#5  main (argc=-1814981296, argv=0xffbefbe4)
    at /nile.build/botcazou/gcc-head/src/gcc/gcov.c:423

In fact the array is empty:

(gdb) p n_names
$1 = 0
(gdb) p names
$2 = (name_map_t *) 0x0

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[nathan at acm dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #6 from Nathan Sidwell <nathan at acm dot org> 2011-11-24 21:36:20 UTC ---
On 11/24/11 19:46, ebotcazou at gcc dot gnu.org wrote:

> In fact the array is empty:
>
> (gdb) p n_names
> $1 = 0
> (gdb) p names
> $2 = (name_map_t *) 0x0

d'oh!  A fix will be right up.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-11-24 21:54:15 UTC ---
> d'oh!  A fix will be right up.

Thanks.  I confirm that adding if (n_names > 0) in the right places works fine.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[nathan at acm dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #8 from Nathan Sidwell <nathan at acm dot org> 2011-11-24 22:12:11 UTC ---
On 11/24/11 21:54, ebotcazou at gcc dot gnu.org wrote:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297
>
> --- Comment #7 from Eric Botcazou<ebotcazou at gcc dot gnu.org>  2011-11-24 21:54:15 UTC ---
>> d'oh!  A fix will be right up.
>
> Thanks.  I confirm that adding if (n_names>  0) in the right places works fine.

Hm, can you try the attached patch?  It avoids passing a null pointer, which is 
not permitted.  Passing zero as nmemb is permitted (7.20.5 para 1 of c99).  So 
I'm a little puzzled as to why bsearch managed to call the comparison function 
at all.  I'd like to understand if we're dealing with a weird, but legal, 
implementation, or one that's got a bug.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-11-24 23:31:06 UTC ---
> Hm, can you try the attached patch?  It avoids passing a null pointer, which is 
> not permitted.  Passing zero as nmemb is permitted (7.20.5 para 1 of c99).  So 
> I'm a little puzzled as to why bsearch managed to call the comparison function 
> at all.  I'd like to understand if we're dealing with a weird, but legal, 
> implementation, or one that's got a bug.

The patch solves the problem for me.

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #10 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-11-25 10:20:28 UTC ---
> The patch solves the problem for me.

Same on i386-pc-solaris2.8 (all gcc.misc-tests/gcov and g++.dg/gcov
tests) and alpha-dec-osf5.1b (with one exception:

Executing on host: /var/gcc/regression/trunk/5.1b-gcc/build/gcc/xgcc
-B/var/gcc/
regression/trunk/5.1b-gcc/build/gcc/
/vol/gcc/src/hg/trunk/local/gcc/testsuite/g
cc.misc-tests/gcov-14.c    -O2 -fprofile-arcs -ftest-coverage  -lm   -o
./gcov-1
4.exe    (timeout = 300)
Foo^M^M
collect2: error: ld returned 1 exit status^M^M
compiler exited with status 1
output is:
Foo^M^M
collect2: error: ld returned 1 exit status^M^M

FAIL: gcc.misc-tests/gcov-14.c (test for excess errors)
Excess errors:
Foo

WARNING: gcc.misc-tests/gcov-14.c compilation failed to produce executable

This happens because ECOFF doesn't support weakrefs,
cf. config/weakref.m4 at the toplevel.

    Rainer

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[nathan at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

--- Comment #11 from Nathan Sidwell <nathan at gcc dot gnu.org> 2011-11-26 21:44:26 UTC ---
Author: nathan
Date: Sat Nov 26 21:44:24 2011
New Revision: 181745

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=181745
Log:
    PR gcov-profile/51297
    * gcov.c (main): Allocate initial names and sources arrays.
    (find_source): Don't check for null name or source arrays here.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/gcov.c

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[nathan at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

Nathan Sidwell <nathan at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |nathan at gcc dot gnu.org
                   |gnu.org                     |

[Bug gcov-profile/51297] [4.7 regression] Many gcov tests FAIL on Tru64, Solaris 8 and 9

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51297

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

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

--- Comment #12 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-12-06 14:10:02 UTC ---
Fixed.