Hi all y'all

I am the consultant helping Aaditya compile deal.II on NERSC's Cori system. 
Let me start by trying to address Wolfgang's questions:

> that would help, but the question is why that is necessary

Good question! I've never had to do this myself (the cray compiler wrappers 
usually take care of things like this). I had a look around -- and a chat 
with some of my colleagues -- and our working theory is that:
1) the cmake script is interfering with the cray compiler wrappers
2) the cmake script is adding libraries that it's detecting in the 
environment

> What does this library do

AtpSigHandler is part of Cray's ATP support -- Abnormal Termination 
Processing (ATP) is a system that monitors Cray system user applications. 
(taken from `man atp_intro`) 

> If this kind of change is necessary for *every* project you want to 
install there, then the system setup is somehow broken 

I think if there is something broken with Cori, loads of other users would 
run into the same problem -- that doesn't mean that everything's right 
either, for example: one of the libraries used by deal.II might be broken. 
If there is a way find out where deal.II's build system gets its link-time 
flags (i.e. is there a verbose mode? can be get more information about how 
cmake detects its environment?)

> so I suspect that there is a reason why the compiler requires linking 
with this project, but I don't know why

This is the first build where we needed to add this library explicitly. And 
looking at past tickets, I can only find a hand full of occurrences of this 
error.

The ATP module is loaded by default (and the cray compiler wrappers link to 
it automatically), so I am also confused why this step is necessary. The 
only things that I can think of are listed above. I agree with them that it 
makes sense to try and debug why this and find out why their build needs 
something special. I was hoping that you (i.e. the deal.II developers) 
would know that this is an issue with Cray systems already. Alas, it seems 
that this is a new problem for you also.

But I am available to work with you directly -- however I cannot debug this 
on my own (as the cmake setup is pretty huge!).

Cheers, 
Johannes

On Monday, August 24, 2020 at 4:15:50 PM UTC-7 Wolfgang Bangerth wrote:

> On 8/24/20 3:40 PM, Aaditya Lakshmanan wrote:
> > 
> > I am not sure what the above means. I contacted support at NERSC for 
> help and 
> > he responded as follows :
> > 
> -----------------------------------------------------------------------------------------------------
> > The problem is that somewhere in the cmake files it's missing:
> > ```
> > find_library(ATP_SIGHANDLER_LIBRARY NAMES AtpSigHandler)
> > ```
> > and then you need to add
> > ```
> > target_link_libraries(<Target> ${ATP_SIGHANDLER_LIBRARY})
> > ```
> > to the file that specifies the link step for `libdeal_II.g.so`
> > 
> > I had a look at the cmake files, but it's a pretty complex setup. I 
> think it 
> > might be best to reach out to the developers again, and get them to 
> > incorporate those changes for cray systems.
> > 
> -----------------------------------------------------------------------------------------------------
> > 
>
> Aaditya -- that would help, but the question is why that is necessary. 
> What 
> does this library do, and why do we need to link with it? If this kind of 
> change is necessary for *every* project you want to install there, then 
> the 
> system setup is somehow broken -- so I suspect that there is a reason why 
> the 
> compiler requires linking with this project, but I don't know why. Can you 
> ask 
> you system support why this is necessary and why the compiler doesn't link 
> these libraries in itself?
>
> Best
> W.
>
> -- 
> ------------------------------------------------------------------------
> Wolfgang Bangerth email: bang...@colostate.edu
> www: http://www.math.colostate.edu/~bangerth/
>
>

-- 
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see 
https://groups.google.com/d/forum/dealii?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to dealii+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/dealii/d5190bc3-753e-4cd8-9255-211c6ce4d63fn%40googlegroups.com.

Reply via email to