Wolfgang,
yes, it is more mild than to be considered as a bug, but code is confusing and implies bug at least for someone that do not read doc carefully, i did not read documentation and just check code in which one expect what i reported is as a bug. Though it is rarely needed, but it seems to be a flaw, e.g. assume we have two scalers, one with Dirichlet BC and one with non-homogeneous Neumann bc, then user have to use two distinct calls of error estimator. but why leave this issue in code, (at least) you can fix it very very easily (just by few minutes), (if you do not like to pass vector of BC objects:) for this you can assign one prototype of error estimator without any BC object for Dirichlet and homogeneous neumann BCs, it can takes a vector of fields. And another protopype for non-homogeneous neumann bc that takes just one field (either scaler or multi-component) and its related BC object. Cheers RT On Tue, Dec 2, 2008 at 5:26 PM, Wolfgang Bangerth <[EMAIL PROTECTED]>wrote: > > > e.g., in "error_estimator.cc" line 1322 we consider single component case > > but can be multi-field, vector "g" stor one scaler (not vector) per QP > > (quadrature ponit), when we take bc, we just take one bc for all field, > and > > as you can see in line 1331, we loop over "n_solution_vectors" and exert > > > > per_thread_data.phi[n][point][0] -= g[point] > > > > actually it should be something like: > > > > per_thread_data.phi[n][point][0] -= g[n][point]; > > > > i.e., do not contrast between various "n" > > > > in the same manner we have this scenario for multi-component vector > fields > > around line 1333. > > Ah, I see. Yes, we use the same boundary conditions for all solution > vectors. > I wouldn't call it a bug (the code doesn't do anything different from what > the documentation says) but we could indeed improve the documentation and > say > explicitly that this is the case. > > W. > > ------------------------------------------------------------------------- > Wolfgang Bangerth email: [EMAIL PROTECTED] > www: http://www.math.tamu.edu/~bangerth/ > >
_______________________________________________
