[petsc-dev] [petsc-checkbuilds] PETSc blame digest (next) 2016-04-08

Tue Apr 12 23:15:50 CDT 2016

On Tue, Apr 12, 2016 at 10:42:26PM -0500, Satish Balay wrote:
> Ok - I see the bug in this code.
> 
> static PetscErrorCode DMPlexReferenceTreeRestoreChildrenMatrices_Injection(DM refTree, Mat inj, PetscScalar ****childrenMats)
> 
> Since this funciton is defined as 'static' the compiler is doing
> 'inter procedure analysis' across the calling function with -O [this
> warning comes up only with -O build]
> 
> 1. plextree.c:3506 has the following variable declared [uninitialized]
>   PetscScalar    ***childrenMats;
> 
> 2. the next usage of this variable is: plextree.c:3846
>   ierr = DMPlexReferenceTreeGetChildrenMatrices_Injection(refTree,injRef,&childrenMats);CHKERRQ(ierr);
> 
> 3. Within this function - the first usage of this variable is plextree.c:L3453
>   refPointFieldMats = *childrenMats;
> 
> i.e access/use of uninitialized variable.
> 
> The following supresses the warning - but there is clearly some lurking bug..

A bug in the code, or the analyzer?  Like I said, there is a
conditional-free path that leads to plextree.c:L3436 (was 3437)---

  *childrenMats = refPointFieldMats;

---in the Get function before the line in the Restore function that's
triggering the warning.  Maybe the (shorter) Restore function is being
inlined while the Get function isn't, I don't know, but I'm fairly
confident this is a false positive. 

Furthermore, this code is covered by tests in next, but looking at the
dashboard I don't see any warnings from the valgrind build.

> 
> diff --git a/src/dm/impls/plex/plextree.c b/src/dm/impls/plex/plextree.c
> index 2ee02d0..e51f704 100644
> --- a/src/dm/impls/plex/plextree.c
> +++ b/src/dm/impls/plex/plextree.c
> @@ -3503,7 +3503,7 @@ PetscErrorCode DMPlexComputeInjectorTree(DM coarse, DM fine, PetscSF coarseToFin
>    PetscLayout    rowMap, colMap;
>    PetscInt       rowStart, rowEnd, colStart, colEnd, *nnzD, *nnzO;
>    PetscObject    injRefObj;
> -  PetscScalar    ***childrenMats;
> +  PetscScalar    ***childrenMats=0;
>    PetscErrorCode ierr;
>  
>    PetscFunctionBegin;
> 
> And this was introduced by http://bitbucket.org/petsc/petsc/commits/f30e825d
> 
> Satish
> 
> On Fri, 8 Apr 2016, Barry Smith wrote:
> 
> > 
> >   Toby,
> > 
> >    We'll be filtering out these false positives. It is just taking us a couple of days to determine that needs to be filtered.
> > 
> >   Barry
> > 
> > > On Apr 8, 2016, at 2:36 PM, Tobin Isaac <tisaac at uchicago.edu> wrote:
> > > 
> > > 
> > > I spent a couple minutes verifying that there is no code path where
> > > that variable is not initialized on line 3437.
> > > 
> > > -Wmaybe-uninitialized seems to generate a lot of false positives.
> > > Coupled with the blame digest's poor ability to assign blame, this
> > > proactive approach seems to waste more time than a reactive
> > > SEGV->valgrind->fix approach.
> > > 
> > > Cheers,
> > >  Toby
> > > 
> > > On Fri, Apr 08, 2016 at 02:12:22PM -0500, Satish Balay wrote:
> > >> This should have been sent to Toby.
> > >> 
> > >> I don't really understand the issue here. Its triggered by --with-debugging=0. I'm
> > >> not sure why these compilers are pointing to petscsys.h. I get slightly different message
> > >> on my laptop.
> > >> 
> > >> mpicc -c -fPIC  -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -fvisibility=hidden -g -O   -I/home/balay/petsc/include -I/home/balay/petsc/arch-linux2-c-opt/include -I/home/balay/soft/mpich-3.1.4/include    -MMD -MP /home/balay/petsc/src/dm/impls/plex/plextree.c -o arch-linux2-c-opt/obj/src/dm/impls/plex/plextree.o
> > >> /home/balay/petsc/src/dm/impls/plex/plextree.c: In function ‘DMPlexComputeInjectorTree’:
> > >> /home/balay/petsc/src/dm/impls/plex/plextree.c:3484:11: warning: ‘childrenMats’ may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>   ierr = PetscFree(refPointFieldMats);CHKERRQ(ierr);
> > >>           ^
> > >> /home/balay/petsc/src/dm/impls/plex/plextree.c:3507:18: note: ‘childrenMats’ was declared here
> > >>   PetscScalar    ***childrenMats;
> > >>                  ^
> > >> 
> > >> Satish
> > >> 
> > >> On Fri, 8 Apr 2016, PETSc checkBuilds wrote:
> > >> 
> > >>> 
> > >>> 
> > >>> Dear PETSc developer,
> > >>> 
> > >>> This email contains listings of contributions attributed to you by
> > >>> `git blame` that caused compiler errors or warnings in PETSc automated
> > >>> testing.  Follow the links to see the full log files. Please attempt to fix
> > >>> the issues promptly or let us know at petsc-dev at mcs.anl.gov if you are unable
> > >>> to resolve the issues.
> > >>> 
> > >>> Thanks,
> > >>>  The PETSc development team
> > >>> 
> > >>> ----
> > >>> 
> > >>> warnings attributed to commit https://bitbucket.org/petsc/petsc/commits/49d7da5
> > >>> PetscMalloc: allow ptr=malloc(0) and free(ptr)
> > >>> 
> > >>>  include/petscsys.h:1157
> > >>>    [http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/08/build_next_arch-freebsd-cxx-pkgs-opt_wii.log]
> > >>>      /usr/home/balay/petsc.clone-2/include/petscsys.h:1157:88: warning: 'childrenMats' may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>>    [http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/08/build_next_arch-linux-opt-cxx-quad_grind.log]
> > >>>      /sandbox/petsc/petsc.clone-3/include/petscsys.h:1157:88: warning: 'childrenMats' may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>>    [http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/08/build_next_arch-freebsd-pkgs-opt_wii.log]
> > >>>      /usr/home/balay/petsc.clone/include/petscsys.h:1157:39: warning: 'childrenMats' may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>>    [http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/08/build_next_arch-linux-pkgs-latest_cg.log]
> > >>>      /sandbox/petsc/petsc.clone-4/include/petscsys.h:1157:25: warning: 'childrenMats' may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>>    [http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/08/build_next_arch-linux-pkgs-opt_crank.log]
> > >>>      /sandbox/petsc/petsc.clone/include/petscsys.h:1157:25: warning: 'childrenMats' may be used uninitialized in this function [-Wmaybe-uninitialized]
> > >>> 
> > >>> ----
> > >>> To opt-out from receiving these messages - send a request to petsc-dev at mcs.anl.gov.
> > >>> 
> > > 
> > 
> > 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20160412/caccfdb5/attachment.sig>