Thread: [Libmesh-users] Bug in system_of_equations_ex8

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi,

Can you confirm the output of the example system_of_equations_ex8 is as
expected?
I see that the constrained face has zero displacement, whereas the
neighboring nodes are ok.

Thanks,
Renato

On Tue, Aug 31, 2021 at 8:11 AM Renato Poli <re...@gm...> wrote:

> Hi,
>
> Can you confirm the output of the example system_of_equations_ex8 is as
> expected?
> I see that the constrained face has zero displacement, whereas the
> neighboring nodes are ok.
>

Hi Renato,

Yes, I see the same issue as you with this example. I have opened a libmesh
ticket regarding this:

https://github.com/libMesh/libmesh/issues/3000

FYI: we are phasing out the mailing list since not many people use it and
it doesn't allow attachments or easy linking with Issues/PRs/branches. The
plan is to try GitHub Discussions (
https://github.com/libMesh/libmesh/discussions) instead for things that
aren't quite formal Issues or PRs yet.

Thanks,
John

Yes, you're right, there is a bug. We need to add:

> system.get_dof_map().enforce_constraints_exactly(system);

before:

libMesh::out << "Computing stresses..." << std::endl;

This imposes the non-zero Dirichlet constraints on the solution, which is
what you need to do for plotting purposes. Note that this has no effect on
the solve itself, it's just required for correct visualization.

I think we also need to add:

> system.update();

in the same place as well so that we also update
system.current_local_solution, since that is what is used in the stress
calculation. I tried that but the stress still seems a bit strange on the
boundary, so not sure what is causing that. But anyway, my suggestion would
be to update this example to have two new lines:

system.get_dof_map().enforce_constraints_exactly(system);
> system.update();

Regards,
David

On Tue, Aug 31, 2021 at 9:11 AM Renato Poli <re...@gm...> wrote:

> Hi,
>
> Can you confirm the output of the example system_of_equations_ex8 is as
> expected?
> I see that the constrained face has zero displacement, whereas the
> neighboring nodes are ok.
>
> Thanks,
> Renato
>
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>

Regarding the GitHub discussions, it does not allow the general public 
to send messages, and it did not have an option for attachments when I 
tried it. May be these would help?
- https://www.bravenet.com/messageforums
- https://www.forumotion.com/
- https://createaforum.com/
- https://createaforum.com/
- https://boardhost.com/

Cheers!

disclaimer: I have nothing to do with those websites, I just want 
something which is convenient for everyone (including myself).

On 2021-08-31 13:53, John Peterson wrote:
> On Tue, Aug 31, 2021 at 8:11 AM Renato Poli <re...@gm...> wrote:
> 
>> Hi,
>> 
>> Can you confirm the output of the example system_of_equations_ex8 is 
>> as
>> expected?
>> I see that the constrained face has zero displacement, whereas the
>> neighboring nodes are ok.
>> 
> 
> Hi Renato,
> 
> Yes, I see the same issue as you with this example. I have opened a 
> libmesh
> ticket regarding this:
> 
> https://github.com/libMesh/libmesh/issues/3000
> 
> FYI: we are phasing out the mailing list since not many people use it 
> and
> it doesn't allow attachments or easy linking with Issues/PRs/branches. 
> The
> plan is to try GitHub Discussions (
> https://github.com/libMesh/libmesh/discussions) instead for things that
> aren't quite formal Issues or PRs yet.
> 
> Thanks,
> John
> 
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users

On Tue, Aug 31, 2021 at 10:28 AM edgar <edg...@cr...> wrote:

> Regarding the GitHub discussions, it does not allow the general public
> to send messages, and it did not have an option for attachments when I
> tried it.

Hi edgar,

By "general public" do you just mean people that don't have GitHub
accounts? The Discussions forum allows anyone with a GitHub account to
start a discussion (e.g. https://github.com/libMesh/libmesh/discussions/2994)
and it looks like it allows for the standard file types (e.g. screenshots)
to be attached which is an improvement over the mailing list. In addition,
you can refer to other Issues/PRs and even lines or blocks of code really
easily since everything is integrated.

--
John

> OK, well I suggest you try again, because we have had two users
> successfully start discussions. It's possible that your account was too 
> new
> or not yet confirmed, and that prevented your Discussion from 
> posting...

I was not planning to answer back to this, because someone once told me: 
"Pick your battles". I think it's a good advice. However, I stumbled 
upon some projects recently which morally oblige to do so.

I know that whatever I say will only alienate you further. I cannot 
convince you of anything, nor intend to do it. The only to which I can 
aspire is to express my opinion, and have a nice exchange of ideas. To 
be crystal clear: I know that this is your project, that I have no say 
in it. At most, I can contribute with some code, possibly answer some 
questions and ask other ones.

The thing about the mailing list goes beyond libMesh or any other 
project. I came across sourcehut recently, and looked for ways to 
migrate there, because they have a way to contribute without 
registration, and they also have mailing lists. In my quest (HAHA... as 
if), I found that the reasons of those who are doing it resonate with my 
beliefs.

There is a point to free software well beyond open source. I will keep 
using the mailing list, and who knows, may be GitHub decides to let me 
communicate with you. May be having the mailing list is too cumbersome. 
I am sure that migrating may be difficult (otherwise I would offer 
myself to do it). Again, it is your project, and I was very wishful to 
use it. I hope that I still can, even without a mailing list.

Thank you, John. You have always been kind.

> Hi edgar,
> 
> By "general public" do you just mean people that don't have GitHub
> accounts? The Discussions forum allows anyone with a GitHub account to
> start a discussion (e.g. 
> https://github.com/libMesh/libmesh/discussions/2994)
> and it looks like it allows for the standard file types (e.g. 
> screenshots)
> to be attached which is an improvement over the mailing list. In 
> addition,
> you can refer to other Issues/PRs and even lines or blocks of code 
> really
> easily since everything is integrated.
> 
> --
> John

Nope, by "general public" I mean anyone with or without a GitHub 
account. When I created an account just to publish my question, it did 
not publish it. When I need a link, I do copy-paste, most of the time, 
which is really convenient. I am sure that a dedicated site for forums 
has more convenient functionalities as well. Not to mention the 
simplicity of having a mailing list, which will likely survive any 
private interest.

On Tue, Aug 31, 2021 at 11:08 AM edgar <edg...@cr...> wrote:

> > Hi edgar,
> >
> > By "general public" do you just mean people that don't have GitHub
> > accounts? The Discussions forum allows anyone with a GitHub account to
> > start a discussion (e.g.
> > https://github.com/libMesh/libmesh/discussions/2994)
> > and it looks like it allows for the standard file types (e.g.
> > screenshots)
> > to be attached which is an improvement over the mailing list. In
> > addition,
> > you can refer to other Issues/PRs and even lines or blocks of code
> > really
> > easily since everything is integrated.
> >
> > --
> > John
>
> Nope, by "general public" I mean anyone with or without a GitHub
> account. When I created an account just to publish my question, it did
> not publish it.

OK, well I suggest you try again, because we have had two users
successfully start discussions. It's possible that your account was too new
or not yet confirmed, and that prevented your Discussion from posting...

-- 
John