Thanks for the patch. I can add it in. Wrt TSAdaptCheckStage() - there is one more warning - its associated with lines Jed changed - so you didn't see this. CC arch-linux2-c-debug/obj/src/ts/adapt/interface/tsadapt.o /home/balay/petsc/src/ts/adapt/interface/tsadapt.c:639:7: warning: Value stored to 'new_dt' is never read new_dt = dt*adapt->scale_solve_failed; ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1 warning generated. Perhaps the following fix? diff --git a/src/ts/adapt/interface/tsadapt.c b/src/ts/adapt/interface/tsadapt.c index 2193b98..5beecdd 100644 --- a/src/ts/adapt/interface/tsadapt.c +++ b/src/ts/adapt/interface/tsadapt.c @@ -635,8 +635,6 @@ PetscErrorCode TSAdaptCheckStage(TSAdapt adapt,TS ts,PetscReal t,Vec Y,PetscBool ierr = PetscViewerASCIIPrintf(adapt->monitor," TSAdapt '%s': step %3D stage rejected t=%-11g+%10.3e, nonlinear solve failures %D greater than current TS allowed\n",((PetscObject)adapt)->type_name,ts->steps,(double)ts->ptime,dt,ts->num_snes_failures);CHKERRQ(ierr); ierr = PetscViewerASCIISubtractTab(adapt->monitor,((PetscObject)adapt)->tablevel);CHKERRQ(ierr); } - } else { - new_dt = dt*adapt->scale_solve_failed; } } else { ierr = TSFunctionDomainError(ts,t,Y,accept);CHKERRQ(ierr); We just enabled this clang analyzer test - its showing quiet a few things - its giving a huge list of issues to go through.. Satish On Thu, 7 Apr 2016, Pierre Barbier de Reuille wrote:

Hey,

I just received this blame: http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/2016/04/07/build_next_a...

I corrected it by initializing dt and new_dt to 0 in TSAdaptCheckStage. Note that dt cannot be used before is properly initialized using the TSGetTimeStep function.

The result is in the branch PierreBdR/initialized-dt-and-new_dt-in-tsadaptchec-1460047069883 <https://bitbucket.org/PierreBdR/petsc-1/branch/PierreBdR/initialized-dt-and-new_dt-in-tsadaptchec-1460047069883> (sorry for the name, I used the online editor for such a small modification).

Cheers,

Pierre