Entering edit mode
Hello
I am facing this strange issue where installation of a package (GO.db) fails due to a dependency (AnnotationDbi) not having the correct version, even though in fact it is the latest available version. The dependency loads fine and shows properly in sessionInfo(). I should specify that this is on a large compute cluster. I install all my packages in a personal library, to which I point with .libPaths(). I am facing no such issue with over 100 other packages.
Thanks for any advice on how I may solve this problem.
Alex
R version 4.3.1 (2023-06-16) -- "Beagle Scouts"
> .libPaths("/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1")
> library("AnnotationDbi")
Error in library("AnnotationDbi") :
there is no package called 'AnnotationDbi'
> BiocManager::install("AnnotationDbi", lib="/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1")
Bioconductor version 3.18 (BiocManager 1.30.22), R 4.3.1 (2023-06-16)
Installing package(s) 'AnnotationDbi'
trying URL 'https://bioconductor.org/packages/3.18/bioc/src/contrib/AnnotationDbi_1.64.1.tar.gz'
Content type 'application/x-gzip' length 4359564 bytes (4.2 MB)
==================================================
downloaded 4.2 MB
[[[snip]]]
* DONE (AnnotationDbi)
The downloaded source packages are in
'/tmp/RtmpDciMJ7/downloaded_packages'
> library("AnnotationDbi")
Loading required package: stats4
Loading required package: BiocGenerics
Attaching package: 'BiocGenerics'
[[[snip]]]
> BiocManager::install("GO.db", lib="/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1")
Bioconductor version 3.18 (BiocManager 1.30.22), R 4.3.1 (2023-06-16)
Installing package(s) 'GO.db'
trying URL 'https://bioconductor.org/packages/3.18/data/annotation/src/contrib/GO.db_3.18.0.tar.gz'
Content type 'application/x-gzip' length 29013599 bytes (27.7 MB)
==================================================
downloaded 27.7 MB
* installing *source* package 'GO.db' ...
** using staged installation
** R
** inst
** byte-compile and prepare package for lazy loading
Error: package 'AnnotationDbi' 1.60.2 was found, but >= 1.63.2 is required by 'GO.db'
Execution halted
ERROR: lazy loading failed for package 'GO.db'
* removing '/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1/GO.db'
The downloaded source packages are in
'/tmp/RtmpDciMJ7/downloaded_packages'
Warning message:
In install.packages(...) :
installation of package 'GO.db' had non-zero exit status
sessionInfo( )
R version 4.3.1 (2023-06-16)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: CentOS Linux 7 (Core)
Matrix products: default
BLAS/LAPACK: FlexiBLAS IMKL; LAPACK version 3.9.0
locale:
[1] LC_CTYPE=en_CA.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_CA.UTF-8 LC_COLLATE=en_CA.UTF-8
[5] LC_MONETARY=en_CA.UTF-8 LC_MESSAGES=en_CA.UTF-8
[7] LC_PAPER=en_CA.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_CA.UTF-8 LC_IDENTIFICATION=C
time zone: America/Vancouver
tzcode source: system (glibc)
attached base packages:
[1] stats4 stats graphics grDevices utils datasets methods
[8] base
other attached packages:
[1] AnnotationDbi_1.64.1 IRanges_2.36.0 S4Vectors_0.40.2
[4] Biobase_2.62.0 BiocGenerics_0.48.1
loaded via a namespace (and not attached):
[1] crayon_1.5.2 vctrs_0.6.5 httr_1.4.7
[4] cli_3.6.2 rlang_1.1.3 DBI_1.2.1
[7] png_0.1-8 bit_4.0.5 RCurl_1.98-1.14
[10] Biostrings_2.70.1 KEGGREST_1.42.0 bitops_1.0-7
[13] fastmap_1.1.1 GenomeInfoDb_1.38.5 memoise_2.0.1
[16] BiocManager_1.30.22 compiler_4.3.1 RSQLite_2.3.4
[19] blob_1.2.4 XVector_0.42.0 R6_2.5.1
[22] GenomeInfoDbData_1.2.11 tools_4.3.1 bit64_4.0.5
[25] zlibbioc_1.48.0 cachem_1.0.8
My personal library is first in the list
However, I am not sure if I understand your first question, about "upstream of my local path". I do have the same package installed in a path at the same level
"/project/6026769/R/x86_64-pc-linux-gnu-library/4.2.2"and more, all the way down to version 3.6 of R. Could this be what is causing me trouble? That would be unexpected for me, considering that libPaths explicitely indicates to go to a specific folder.What I meant was that the sysadmin might have already installed
AnnotationDbiin the site-wide library dir, and if your.libPathsvariable had the site-wide dir first, then you would be comparing versus that one rather than your local version.Having other libdirs that you use for other versions of R is fine (I have the same setup), so I wouldn't expect that to be a problem. I guess the obvious thing to try is
And see if that picks something up. I wouldn't expect it to, but it's worth a try.
Thanks for helping me with this. Unfortunately (or not?), there doesn't seem to be any problem with Bioc
And if you try to install
GO.dba second time it still fails?Yes, it still fails with the same error message about the version of AnnotationDbi being too low. No matter if I issue
library("AnnotationDbi")ahead of the install call or not.I realize this problem may be caused by some particular configuration of the cluster I am using. I don't have the skills to figure this out myself, and I am not expecting you to troubleshoot this kind of problem. I have asked the cluster tech support team for some help. I was provided with a working but "trivial" solution to load a different R module that comes pre-bundled with Bioconductor. Unfortunately, no clear explanation as to why the GO.db installation does not see my local AnnotationDbi library...
This is a weird one, and might be difficult to track down. These messages
all come from
tools:::.install_packages, which is called 'outside' of the R session that you are installing from. In other words, all packages are installed usingR CMD INSTALL, which is a shell script that calls another instance of the R binary to do the actual installation. When you useBiocManager::installit checks to ensure that the versions all line up, and then passes the correct packages to install toutils::install.packages, which then (among other things) generates the correct arguments forR CMD INSTALLand passes them off to that script usingsystem2, which is a way to make system calls from within R.What that means is you cannot do the obvious thing - run
tools:::.install_packagesunder the debugger - to figure out why the 'wrong' version ofAnnotationDbiis being identified, because it's being run by a different instance of R.Pardon my ignorance, but would it be possible then to pass the location of my R libraries with something like
export R_LIBS=/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1or as a variable in the appropriateR CMDcommand?It already does. If I install
GO.dbunder the debugger (forinstall.packages), and proceed through the function, I end up here:So what is being passed to
system2is the path to the R binary (cmd0), and a set of arguments, which will then create a system call ofwhich is already pointing to the correct library dir.
Could you check that
AnnotationDbiis really installed with the right version in the desired library viainstalled.packages(lib.loc = "/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1")? Is this with a fresh installation without loading anything? This might also cause problems if the package from the system is loaded before you install and try to load the new one in your personal/project libraryI am not sure about your question regarding "fresh installation", as I work on a cluster where I load pre-installed modules for R and its dependencies. However, without loading anything, immediately after starting my R session, I get:
Yes, by fresh installation, I meant without loading anything before doing the installation steps. Now it seems you fulfill the GO.db requirements: package 'AnnotationDbi' 1.63.2 is installed. Can you try again without doing anything else:
BiocManager::install("GO.db", lib = "/project/6026769/R/x86_64-pc-linux-gnu-library/4.3.1"), if it finds some missing dependencies BiocManager should install them.If you still have problems I recommend to create an .Rprofile in the directory where you lunch the R process with a line reporting the libraries it has (just a
print(.libPaths()\nshould be enough but you might want to addprint(packageVersion("AnnotationDbi")) . Maybe the way it loads R interferes with finding the package (which shouldn't happen).and
I am unsure why
packageVersionfails when the .Rprofile script is run at the beginning of the R session, but the command works immediately afterwards.Maybe utils is not loaded yet. It might work if you use utils::packageVersion. But anyway, does this mean that you still cannot install GO.db? Clever idea to put the .libPaths in the R profile, this might help to find the right library when installing packages
and
Notice how:
installcommand. I was not expecting that.If I specify
lib=in theinstallcommand, then it did not complain about the packages to update (they must be in my user library and up-to-date there) and it found other ones needing updating:GOsemSim depends on GO.db, so both fail.
I finally kind of solved it. I also had a .Rprofile file in my home (~) This
~/.Rprofilewas establishing .libPaths to libraries for an earlier version of R (4.2.2) with of course the wrong version ofAnnotationDbiAlthough R was properly sourcing the .Rprofile located in my projects directory at the start of the session, it was somehow reverting to
~/.Rprofilefor the installation. I was not expecting this, and frankly I don't understand.Here is some proof:
I should apologize to James and Lluis for the time wasted on this.
That's why I have a print/message in my .Rprofile: showing when I am running it (it helps debug this kind of problems) ;) Nice to see it solved