Closed BertvanUlft closed 2 years ago
Hi!
I've finally had time to look into this. This was a bug in the install script, thanks for reporting. It should now be fixed (v4.3.0). See https://github.com/Hirlam/obsmon/commit/8d8f55dba7b5130d41d06fc43d42deaa64b29fbf.
I'll close the issue but feel free to make a new report should you continue to face this issue.
Thanks Paulo, had to update versions of three packages, but after that installation ran smoothly!
What do you mean? Did you need to update packages after this update? You shouldn't in principle need to do this after this update, but, if you did, which packages were those and why did you need to update them?
Compilation of three packages failed with the list in git. I changed these to more recent changes, that I also used in my dodgy obsmom-4.2.0 installation:
isoband == 0.2.4 --> 0.2.5
RcppTOML == 0.1.3 --> 0.1.7
testthat == 3.0.2 --> 3.1.1
I haven't stored the errors it gave.
Would you be able to run the install again with the original packages (ans using the --live
option )? It would be good for me to understand why they failed for you, because in my (limited) test environments the install worked fine. I may even update these packages in the repo if I see that it's a general issue.
For context: The reason I fixed these pkg versions is that they are updated quite frequently on CRAN and then things can often start to misbehave as a result. And that's something I don't have control over.
Great, didn't know the --live option. I tried it on a different Fedora 35 system, it gave the same errors (see below, click to expand). Is there a good method to find a consistent, non-conflicting set of package versions?
When running obsmon, and selecting things in the browser, obsmon prints messages like:
Warning in mccollect(jobs = jobs, wait = TRUE) :
1 parallel job did not deliver a result
ERROR [2022-02-10 12:13:25] Error: Failed to retrieve the result of MulticoreFuture (<none>) from the forked worker (on localhost; PID 549786). Post-mortem diagnostic: No process exists with this PID, i.e. the forked localhost worker is no longer alive.
Unhandled promise error: Failed to retrieve the result of MulticoreFuture (<none>) from the forked worker (on localhost; PID 549786). Post-mortem diagnostic: No process exists with this PID, i.e. the forked localhost worker is no longer alive.
Could that be related to these changed versions?
I am facing the same issue as @BertvanUlft on Ubuntu 22.04, when I'll try to deploy obsmon with ansible. I am checking out version 4.3.0, but I still see the error:
Progress 15%, installing R-lib 24/154: stringi (== 1.7.3) ...
"stderr": "Error: installation of package ‘stringi’ had non-zero exit status
Error in file(con, \"r\") : cannot open the connection
Calls: install ... tryCatchOne -> <Anonymous> -> tryCatch -> readLines -> file
In addition: Warning message:
In file(con, \"r\")
cannot open file 'stringi.out': No such file or directory\nExecution halted"
I have tried installing stringi=1.7.3
using devtools before installing obsmon, which succeeds, but the obsmon install script still gives the same error.
For reference I give my R related ansible steps below:
When installing obsmon 4.1.0 & 4.2.0 on Fedora 35 it fails with errors like:
Installing the package stringi in R manually in the directory where obsmon expects it works fine. After then tar.gzipping it in
obsmon-4.2.0/.installer_local_R-lib/compiled_binaries/x86_64-redhat-linux-gnu-Fedora-35/R_4.1
and updating the .installer_pkg_versions.txt it goes a bit further until it crashes in some next package. Having repeated this procedure till a 157 have been installed it seems to work well. I then expected that I could easily re-install obsmon-4.2.0 on another Fedora 35 machine using the modified .installer_pkg_versions.txt file, but it crashes in the same manner in the stringi installation (now 1.7.6). Note theNA: command not found
error. Any ideas what could be causing this?