Library not installed

[kalmSveta]

I am on Ubuntu 18.04.5, I am trying to install CytoML on R 4.0 with BiocManager::install('CytoML') However I get this error:

Error: package or namespace load failed for ‘CytoML’ in dyn.load(file, DLLpath = DLLpath, ...):
 unable to load shared object '/home/kalmsveta/R/x86_64-pc-linux-gnu-library/4.0/00LOCK-CytoML/00new/CytoML/libs/CytoML.so':
  /home/kalmsveta/R/x86_64-pc-linux-gnu-library/4.0/00LOCK-CytoML/00new/CytoML/libs/CytoML.so: undefined symbol: _ZNK5boost10filesystem15directory_entry20m_get_symlink_statusEPNS_6system10error_codeE
Error: loading failed
Execution halted
ERROR: loading failed

Could you help me, please?

[mikejiang]

try to reinstall cytolib to see if it helps.

[kalmSveta]

try to reinstall cytolib to see if it helps.

I have reinstalled cytolib, but installation of CytoML still produces the same error.

[mikejiang]

please save the complete the output of

remotes::install_github("RGLab/cytolib")
remotes::install_github("RGLab/flowWorkspace")
remotes::install_github("RGLab/CytoML")

to a txt file and attach here

[kalmSveta]

The file is attached install.txt

[mikejiang]

What is your BH package version?

 packageVersion("BH")
[1] ‘1.72.0.3’

[kalmSveta]

I have the same version:

> packageVersion("BH")
[1] ‘1.72.0.3’

[mikejiang]

You may have some older version of boost library on your system (e.g. at /usr/include/boost) that doesn't have directory.hpp header , which precedes the header of your BH package.

One option is to remove your system boost lib, but since boost is pretty commonly installed, I've updated it with the more backward compatible call boost::filesystem::symlink_option::recurse, now it should be good. Try remotes::install_github("RGLab/CytoML") again.

[kalmSveta]

I've tried to reinstall CytoML, but still get the same error

[mikejiang]

Your previous error was this

../inst/include/CytoML/workspace_type.hpp:65:101: error: ‘directory_options’ is not a member of ‘fs’
         for(const fs::directory_entry & i: fs::recursive_directory_iterator(fs::path(data_dir), fs::directory_options::follow_directory_symlink))

Now directory_options reference is removed from the code base so I am a little surprised to hear that you were still getting the same error. Can you confirm with the attached installation log? Also, have you tried to look for your system-wise boost library instance and remove it if it is older than 1.72.0.3?

[kalmSveta]

Sorry, I mean the previous error with Error: package or namespace load failed for ‘CytoML’ in dyn.load(file, DLLpath = DLLpath, ...): unable to load shared object '/home/kalmsveta/R/x86_64-pc-linux-gnu-library/4.0/00LOCK-CytoML/00new/CytoML/libs/CytoML.so': /home/kalmsveta/R/x86_64-pc-linux-gnu-library/4.0/00LOCK-CytoML/00new/CytoML/libs/CytoML.so: undefined symbol: _ZNK5boost10filesystem15directory_entry20m_get_symlink_statusEPNS_6system10error_codeE The log is attached: install2.txt

[mikejiang]

symbol _ZNK5boost10filesystem15directory_entry20m_get_symlink_statusEPNS_6system10error_codeE (i.e. boost::filesystem::directory_entry::m_get_symlink_status(boost::system::error_code*) const is not used by either cytolib or CytoML code bases). Not sure why your build is still looking for it.

Since I can't really reproduce your error, I will have to try my best to guess based on your output, here is my linker flags

 ~/rglab/workspace/CytoML$R CMD config SHLIB_CXXLDFLAGS
-shared

while yours seems to have some other unconventional flags (e.g. -Bsymbolic-functions), but I am not sure if they are relevant to your loading error.

Again, try to locate the boost library under your /home/kalmsveta/anaconda3/ (both include and libs), if it is present then try to remove them to see if it helps.

[Taimie48]

My apologies for resurrecting this issue.

I am trying to install CytoML on a jupyterhub based version of RStudio. I am getting the same "undefined symbol" error as above.

We suspect that the error may result from the version of boost to compile CytoML being different from the one installed on the server. May I ask which version was used to compile CytoML or is there anything else I can do befor asking the owner of the server to change their architecture?

[mikejiang]

current version of BH (https://cran.r-project.org/web/packages/BH/index.html) builds ok with cytoML on bioc site, http://bioconductor.org/checkResults/3.15/bioc-LATEST/CytoML/ so make sure your system boost library is consistent with BH (which is at boost 1.78.0 )