OHDSI Home | Forums | Wiki | Github

Hades install issue

General
#1

I verified that my prereqs worked for my install by testing at the following link: https://ohdsi.github.io/Hades/rSetup.html#Verifying_the_installation

The first command works fine but when I run I get an error on install. I have tried to install the cachem in Rstudio without any luck. This may be an issue with Rstudio but I am getting the issue when I follow the install insturctions at
https://ohdsi.github.io/Hades/installingHades.htmlinstall.packages(“remotes”)
remotes::install_github(“OHDSI/CohortMethod”) as well.

  • installing source package ‘cachem’ …
    ** package ‘cachem’ successfully unpacked and MD5 sums checked
    ** using staged installation
    Warning in as.POSIXlt.POSIXct(x, tz) : unknown timezone ‘GMT’
    ** libs
    Warning: R include directory is empty – perhaps need to install R-devel.rpm or similar

*** arch - i386
"C:/rtools40/mingw32/bin/“gcc -I"C:/PROGRA~1/R/R-40~1.3/include” -DNDEBUG -O2 -Wall -std=gnu99 -mfpmath=sse -msse2 -mstackrealign -c cache.c -o cache.o
cache.c:1:10: fatal error: R.h: No such file or directory
#include <R.h>
^~~~~
compilation terminated.
make: *** [C:/PROGRA~1/R/R-40~1.3/etc/i386/Makeconf:222: cache.o] Error 1
ERROR: compilation failed for package ‘cachem’

#2

The cachem package appears to fail installation, which is required for shiny which in turn is required by many HADES packages.

I don’t know why cachem is causing this error. If I try to install the latest version on Windows myself it works fine. This issue has also not been reported on the cachem issue tracker.

I do see that you appear to be using a 32-bit architecture, where the failure appears. Do you really need the 32-bit architecture? We strongly recommend only installed the 64-bit version of R.

#3

Thanks for the reply. The cachem does not install independently which I find odd. I have looked at the the cachem issue tracker and they did not have much information. I will have another user test in our system. I may need to try to do a reinstall R and Rstudio.

I do understand as I watch a video OHDSI recommending not loading 32 bit but our load is institutional so we prefer not to modify the install. I would hope those that code the program take that into account when they code. I do have the ability to load the 64bit only for myself so I may test if that is an issue.

#4

I just tried to run the code with Rgui 64 bit and I am getting the same error. Odd that it still registers trying to install 32 bit even when I run in the 64 but gui.

#5

Not sure the 32-bit architecture is really the issue, but seems a plausible lead to follow. R automatically builds for all installed architectures, even when you’re not using them.

You can try this to force installing the 64-bit ones only:

remotes::install_github("ohdsi/Hades", INSTALL_opts = c("--no-multiarch"))

The reasons I don’t usually recommend this is that from then on, you must always force everything to be 1 architecture.

#6

Thanks I just tried this command line and I am now getting an odd error “cannot open URL ‘https://OHDSI.github.io/drat/bin/windows/contrib/4.0/PACKAGES’” The web site does not come up in a browser so I am thinking something is now wrong on the site. I have reloaded the computer that I am having install issues to see if that would fix some issues. I have not really be able to install any programs from the documentation. I have tried this command remotes::install_github(“OHDSI/CohortMethod”) as well and there seems to always be some dependency that does not load.

#7

Could you provide a bit more context around the error message? It would be good to know where in the process it pops up.

The error message comes from drat, which can’t find the binaries folder, which is expected because there is no binaries folder. Normally, when you run drat it should then revert to the source folder which is there. However, what is odd is that we no longer use drat anymore. Everything now should run via GitHub and the remotes package.

#8

Ran remotes::install_github(“ohdsi/Hades”, INSTALL_opts = c("–no-multiarch")) this morning

package ‘digest’ successfully unpacked and MD5 sums checked
Error: Failed to install ‘Hades’ from GitHub:
Failed to install ‘BigKnn’ from GitHub:
(converted from warning) cannot rename file ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0\file1f342023304c/digest’ to ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0/digest’, reason ‘This operation is supported only when you are connected to the server’

I got this one yesterday and I looked in the directory the file1f342023304c directory does not exist in the folder location. So I attempted to load digest which will load but when the remote script runs it removes the software. It seems like there is an issue with the install script. When I run the program after loaded digest I get this.

package ‘digest’ successfully unpacked and MD5 sums checked
Error: Failed to install ‘Hades’ from GitHub:
Failed to install ‘BigKnn’ from GitHub:
(converted from warning) cannot rename file ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0\file1f3427831573/digest’ to ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0/digest’, reason ‘This operation is supported only when you are connected to the server’

Out of curiosity I tried to run remotes::install_github(“ohdsi/BigKnn”, INSTALL_opts = c("–no-multiarch")) but I get a similar error.

package ‘digest’ successfully unpacked and MD5 sums checked
Error: Failed to install ‘BigKnn’ from GitHub:
(converted from warning) cannot rename file ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0\file1f34339616f4/digest’ to ‘\geisinger.edu\dfs\0003\0001\USERID\My Documents\R\win-library\4.0/digest’, reason ‘This operation is supported only when you are connected to the server’

#9

Sorry, no idea what is going on. Is your R library on a network drive?

#10

Yes our default install is using a network location it uses our my documents location. I did attempt to move the library and working directories to my local drive yesterday. R was still was using a network drive for the temp directory which seemed to be an issue as well. I did not have any luck with installing local so I am going to try to get a test machine and load everything with R by hand.

I am also going to try to see if I can get a Linux machine setup to see if it works better we do not have any rules on the installation of R on Linux.

Thanks for taking time to review this issue two of us have been working on this for days.

#11

Did you resolve this issue?

I am having similar frustration:

Screenshot 2023-04-05 at 7.22.52 PM
Screenshot 2023-04-05 at 7.22.52 PM893×484 68.8 KB

#12

Hi @Sanjay_Udoshi ! It seems you’re having trouble installing rJava. Do you have Java installed? Did you try sudo R CMD javareconf?

1 Like
#13

Sorry, I should have updated my progress here. Yes! Finally that is what I tried and everything worked swimmingly from there.

Is there a way to make these (ohdsi/hades) globally available for all users on the RStudio Server?

#14

I’m not very familiar with RStudio Server. There appear to be multiple ways to provide a set of packages to all users, as described here.

1 Like
t