Hello all,
Just a short e-mail to give some news (work is in progress still).
I believe that I found the origin of the bug discussed in issue 694548. Not
much related to what I suspected initially (hence, this will be the subject of
another thread ️).
It turns out that the bug was a bit "big" from the start, in the sense that I
believe the original writer of the code would have noticed it easily. Hence, it
makes me a bit suspicious… Is my "fix" actually safe to use? Or was the code
meant to be written this way, and my fix will break something else?
Therefore, before sharing my findings I wanted to check if the MWE given in
issue 694548 also fails with the first version of the code (dfec202) to feature
what I believe is an error in the source. If not, then perhaps my "fix" will be
breaking something else…
As of now, I am trying to compile Yade as of commit dfec202, which turned out
to be quite a nightmare… And that is where I am now still… That being said, I
am hopeful to solve it soon. ️ I'll give some more news early next week,
hopefully to close this issue.
Klaus Thoeni, Wed, 24 Feb 2021 11:36:21 +1100
> I will have to look into this. In the meantime, it would be
> excellent if you could share your solution, best is to push
> your changes to your fork on gitlab if you have one.
As it is, it still needs some cleanup before it can be submitted. That's why I
proposed to discuss the general idea first (otherwise, it's not worth the
effort).
But first, let's close 694548. Since it is a more localized issue, it helps me
build confidence in interacting with an existing opensource project. So not a
waste of time! ️
Then I'll create a new thread/discussion for the original topic I wanted to
discuss.
Janek Kozicki, Wed, 24 Feb 2021 12:55:15 +0100
> reformat with the same .clang-format file your old branch (*).
Thanks for the tip Janek ️
Bruno Chareyre, Fri, 26 Feb 2021 15:00:53 +0100
> I'm eager to understand which problem you found. :)
Thanks for the warm welcome! ️
And I believe I already have an account with the required rights and everything
(we communicated last year). But let's check that when the moment comes…
Gaël
De: Yade-dev
en nombre de Bruno Chareyre
Enviado: viernes, 26 de febrero de 2021 11:00
Para: yade-dev@lists.launchpad.net
Asunto: Re: [Yade-dev] A suggestion/contribution regarding bug of Yade-users
#694548
Hi Gaël,
Thank you for your email. And no you are not colliding with my work since I
have it in mind but didn't start anything yet.
I would suggest you create a gitlab account if you don't have one so we can add
you as yade contributor.
You can then push your change to a branch. Something like (assuming you
modified the sources directly in local master branch):
git push origin master:PFacetFix
I'm eager to understand which problem you found. :)
Bruno
On 23/02/2021 21:59, Gael Lorieul wrote:
Hi ️
It has been a long time…
I have come accross thread #694548 on the "Yade-users" mailing list (from
January 2021), which concerned unphysical collisions between PFacets (the
dynamics of the rebound does not correspond to what one's physical/common sense
would expect).
I wanted to share that I encountered a very similar issue but with
gridConnection objects (I work with cylinders), and that I found a solution.
The latter should be easily extendible both to PFacet/PFacet and
PFacet/gridConnection collisions (although I only tested
gridConnection/gridConnection collisions). I am not sure how elegant or
computationally efficient it is, but it has given me good results on the (very
few) simulations I have tested it on.
I initially wanted to test my solution on full-scale simulations before
submiting it to the Yade code base, so as to present a code that is robust and
mature. But as it has been almost a full year since then (the company I work
for gave me assignments other than DEM), I feel that perhaps I should reach out
for other ways to share this information…
This implies that although I am quite confident of my fix, I am not 100%
certain either…
Perhaps a first step would be for me to present what I have found, and how my
fix works? I am not sure as to where to present it: This mailing list? Thread
#694548? You tell me ️
Gaël
PS: I have started to re-implement my solution in yade e07e530 (26 Nov 2020)
(my previous implementation was based on yade 1b4ae97 (8 Feb 2020)). Hopefully
the work I do on e07e530 will not be difficult to port to the latest yade
versions… (It seems the yade project started using clang-format (or similar
tool) sometime between e07e530 and 1b4ae97 so `diff` outputs half the file: not
very helpful… ️)
PS2: Could thread #695558 be also concerned by this same uncorrect physical
behavior?
PS3: I hope that I am not colliding with Bruno's work on the subject (according
to Bruno's e-mail of Tue, 12 Jan 2021 07:50:51)
