<p dir="ltr">Ummmm. OK. We should remove custom -fd arguments from every example first and use -snes_fd_coloring instead. The example problems should work with that and if they do not then we fix it. The context should always be owned elsewhere so if we manually unset it we should be fine.</p>
<p dir="ltr">- Peter </p>
<div class="gmail_quote">On Aug 15, 2012 8:31 AM, "Jed Brown" <<a href="mailto:jedbrown@mcs.anl.gov">jedbrown@mcs.anl.gov</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="gmail_quote">On Wed, Aug 15, 2012 at 6:41 AM, Barry Smith <span dir="ltr"><<a href="mailto:bsmith@mcs.anl.gov" target="_blank">bsmith@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I think it uses null in order to delay computing the matrix coloring etc from the time of SNESSetJacobian() to a later time. I guess what could happen is it creates a n "empty" fdcoloring beast and passes that into SNESSetJacobian() and then fills it up later when it has the information it needs.</blockquote>
</div><br><div>Right, I think I would prefer that. A side-effect is that the caller setting coloring is now responsible for ownership of the context. but it is for everything else anyway.</div><div><br></div><div>Lisandro asked for destructors for those contexts, but I'm not sure we need a separate one for every context slot. Would one per SNES be enough?</div>
</blockquote></div>