[Users] Problem installing the ETK

Roland Haas rhaas at phas.ubc.ca
Thu Jun 19 18:02:42 CDT 2025


Hello Konstantin,

"DSO missing from command line" usually happens if you are not listing
all libraries. Note that with static libraries order matters and eg for
libsiloh5.a which requires HDF5 functions you need to have them like
this:

-lsiloh5 -lhdf5

ie the library that provides the functions last.

Yours,
Roland

> [CAUTION: Non-UBC Email]
> 
> Hello José,
> 
> 
> I did manage to compile it on wednesday with my own OptionsList, just
> had to call make after the compilation stopped due to the error I've
> described, then it compiled without problems.
> I the tried it again with your OptionsList and it did not fail with
> that error, but at the very end of the process I got a new error
> involving hdf5m reading
> 
> 
> /usr/bin/ld:
> /home/KK/Programieren/ETK/Cactus/configs/sim/scratch/external/Silo/lib/libsiloh5.a(H5FDsilo.o):
> undefined reference to symbol 'H5FDregister'
> /usr/bin/ld: /lib/libhdf5.so.310: error adding symbols: DSO missing
> from command line
> collect2: Fehler: ld gab 1 als Ende-Status zurück
> 
> 
> which did not pop up with my config. The only difference in how we
> handle hdf5 is that I explicitly tell it where the /Lib and Include
> directories are. Maybe there is some effect due to the different
> compiler flags, but I am not that experienced with using them, so I
> have no clue where it would stem from.
> I also did not get any core dumps yet, but I have also only run the
> WaveToy thorn twice, so I'll have to see what happens in the future.
> Anyways, thank you very much for your help.
> 
> 
> With best regards
> 
> 
> Konstantin Kaufhold
> 
> Am Mittwoch, den 04-06-2025 um 17:18 schrieb José Ferreira:
> 
> 
>  
> 
> Hello Konstantin,
> 
> 
> 
> 
> 
> 
> 
> 
> I've sucessfully compiled the toolkit on an Arch Linux machine with
> the option files that I'm sending as an attachment.
> 
> 
> 
> The dependencies are all listed in the option file inside, and are all
> available and up-to-date in the official repositories.
> 
> 
> 
> 
> 
> 
> 
> 
> There is one small caveat: one of the updates of HDF5 in the Arch
> repositories somehow broke the Toolkit compilation, now it fails to
> find HDF5.
> 
> 
> 
> 
> I wanted to look into it and figure out whats going on, but I decided
> to be lazy instead and just downgrade HDF5.
> 
> 
> 
> Version 1.14.5-1 in repositories works.
> 
> 
> 
> Since I don't have any other package depending on HDF5, this works
> fine for me.
> 
> 
> 
> If you figure out what happened, please do let me know.
> 
> 
> 
> 
> 
> 
> 
> 
> Also, when I started using the toolkit, I was getting random core
> dumps on Arch and Arch based distros.
> 
> 
> 
> This hasn't happened to me in many months now, so it was probably
> something to do with a specific version of one of the libraries.
> 
> 
> 
> Either way, if that does happen to you, here's the bug report that I
> made:
> https://bitbucket.org/einsteintoolkit/tickets/issues/2774/openmp-heisenbug-with-default-thornfile
> 
> 
> 
> 
> 
> 
> 
> 
> Best,
> 
> 
> 
> José Ferreira
> 
> 
> 
> 
> 
> 
> 
> 
> On 04/06/25 13:02, Kaufhold, Konstantin wrote:
>  
> 
>  
> Hi, 
> 
>  
> I am currently trying to install the ETK to my personal laptop for a
> project.  
> I am using an ArchLinux system with linux 6.14.9. The processor in my
> laptop is an AMD Ryzen 7 PRO 8840U. 
> While trying to compile the toolkit I get the error
>  
> 
>  
> /usr/include/c++/15.1.1/cstdlib:83:15: schwerwiegender Fehler:
> stdlib.h: Datei oder Verzeichnis nicht gefunden
>    83 | #include_next 
>  
> 
>  
> which then causes the compilation to fail. 
> The only version of the bug I found, which I could reproduce locally,
> was using "gcc -isystem /usr/include" as described in an bug report
> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70129) from 2016 for
> gcc. I tried compiling a small Hello World script written in C++ with
> "gcc", "gcc -I" and gcc -isystem" and the last option gave me the same
> error, the rest worked just fine. 
> But I'm not sure if that is the real reason as other OptionFiles
> indicate, that gcc can and has been successfully used to compile the
> ETK and I have not seen the "-isystem" glag invoked anywhere, so I'm
> guessing the the error is on my side, although I have no clue where
> exactly. 
> I have attached the make.log file and my OptionsList, as I had to
> explicitly give the ETK the HDF folders and then decided to do it for
> most external libraries I saw in the list and have locally on my
> system. 
> I appreciate any ideas or advice on fixing this problem and thank you
> in advance for taking the time to look over my problem. 
> 
>  
> With best regards 
> 
>  
> Konstantin Kaufhold 
>  _______________________________________________ Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users


-- 
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.einsteintoolkit.org/pipermail/users/attachments/20250619/730eee15/attachment.sig>


More information about the Users mailing list