Function power() causing R session to crash

Question

Function power() causing R session to crash

eduardohorta opened this issue 4 years ago · comments

Entering, e.g., power(2,3) will deliver the following error message:

 *** caught illegal operation ***
address 0x110f29f80, cause 'illegal opcode'

A minimal code example that reproduces the bug:

library(CVXR)
power(2,3)

Output

> library(CVXR)

Attaching package: ‘CVXR’

The following object is masked from ‘package:stats’:

    power

> power(2,3)

 *** caught illegal operation ***
address 0x104d6bf80, cause 'illegal opcode'

Traceback:
 1: gmp::as.bigq(p)
 2: pow_high(p)
 3: .local(.Object, ...)
 4: initialize(value, ...)
 5: initialize(value, ...)
 6: new("Power", ...)
 7: .Power(x = x, p = p, max_denom = max_denom)
 8: power(2, 3)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection:

Version

> sessionInfo()
R version 4.0.2 (2020-06-22)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Catalina 10.15.6

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRblas.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib

locale:
[1] C/UTF-8/C/C/C/C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] CVXR_1.0-8

loaded via a namespace (and not attached):
 [1] bit_4.0.4       compiler_4.0.2  R6_2.4.1        Matrix_1.2-18  
 [5] gmp_0.6-0       Rcpp_1.0.5      bit64_4.0.5     grid_4.0.2     
 [9] Rmpfr_0.8-1     lattice_0.20-41

Balasubramanian Narasimhan · Answer 1 · Mon Sep 21 2020 22:50:20 GMT+0800 (China Standard Time)

This is not a reproducible bug and your session info shows that CVXR is not attached.

> library(CVXR)

Attaching package: ‘CVXR’

The following object is masked from ‘package:stats’:

    power

> power(2,3)
Power(2, 3, c("1/3", "2/3"))> 
> sessionInfo()
R version 4.0.2 (2020-06-22)
Platform: x86_64-apple-darwin19.5.0 (64-bit)
Running under: macOS Catalina 10.15.6

Matrix products: default
BLAS/LAPACK: /usr/local/Cellar/openblas/0.3.10_1/lib/libopenblasp-r0.3.10.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] CVXR_1.0-8

loaded via a namespace (and not attached):
 [1] bit_1.1-15.2    compiler_4.0.2  R6_2.4.1        Matrix_1.2-18  
 [5] gmp_0.6-0       Rcpp_1.0.5      bit64_0.9-7     grid_4.0.2     
 [9] Rmpfr_0.8-1     lattice_0.20-41

Eduardo Horta · Answer 2 · Tue Sep 22 2020 01:22:15 GMT+0800 (China Standard Time)

This is not a reproducible bug and your session info shows that CVXR is not attached.

Sorry @bnaras, I forgot to copy the full output. The output is

> library(CVXR)

Attaching package: ‘CVXR’

The following object is masked from ‘package:stats’:

    power

> power(2,3)

 *** caught illegal operation ***
address 0x104d6bf80, cause 'illegal opcode'

Traceback:
 1: gmp::as.bigq(p)
 2: pow_high(p)
 3: .local(.Object, ...)
 4: initialize(value, ...)
 5: initialize(value, ...)
 6: new("Power", ...)
 7: .Power(x = x, p = p, max_denom = max_denom)
 8: power(2, 3)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection:

Notice that CVXR appears under other attached packages in the session info.

Curiously, this same example does work in a rstudio.cloud session.

Balasubramanian Narasimhan · Answer 3 · Tue Sep 22 2020 03:49:30 GMT+0800 (China Standard Time)

So this indicates there is something peculiar to your R installation since it works for my R session too. I suspect it is your gmp R library or the libgmp that it is linking to. It is easiest to uninstall gmp (both R and the library) and reinstall.

Eduardo Horta · Answer 4 · Wed Sep 23 2020 02:31:47 GMT+0800 (China Standard Time)

@bnaras Thank you for the feedback. Unfortunately, uninstalling and reinstalling R and gmp didn't work.

p-palacios · Answer 5 · Fri Sep 25 2020 13:06:07 GMT+0800 (China Standard Time)

I have exactly the same issue. My R version is 4.0.2, mac Catalina 10.15.6 and CVXR 1.0-8. I have tried on Rstudio and from the terminal. The issue appears when i run either power(2,3)
or when i was trying with the vignette line: objective <- Minimize(sum((Y - X %*% betaHat)^2))

I am new with this package, any help will be great.

Balasubramanian Narasimhan · Answer 6 · Sun Sep 27 2020 01:11:25 GMT+0800 (China Standard Time)

This could be a problem with your settings on Catalina which is far more picky about libraries. You may have to allow full disk access to /usr/bin/ruby.

Eduardo Horta · Answer 7 · Sun Sep 27 2020 21:21:33 GMT+0800 (China Standard Time)

@bnaras Unfortunately providing full disk access to /usr/bin/ruby (and also /usr/bin/local/R) did not solve the issue.

@p-palacios Indeed, I first bumped into this error when trying to optimize a quadratic loss as you did. From rStudio there was no error message: the session simply crashed. After investigating a little bit (starting R through the Terminal) I came up with this minimal (non-)working example. If you have urgency in using the CVXR package:

Curiously, this same example does work in a rstudio.cloud session.

Balasubramanian Narasimhan · Answer 8 · Mon Sep 28 2020 07:00:08 GMT+0800 (China Standard Time)

So far we don't have a reproducible example. In a brand new R session, can you run the statement

gmp::as.bigq(3)

Can you do the same in a new Rstudio session?

p-palacios · Answer 9 · Mon Sep 28 2020 07:51:11 GMT+0800 (China Standard Time)

Dear @bnaras gmp::as.bigq(3) does not work with RStudio, last version, or in the terminal, unfortunately. I have already tried the access to /usr/bin/ruby. It didn't work.

@eduardohorta thanks for the suggestion of using RStudio cloud.

Balasubramanian Narasimhan · Answer 10 · Mon Sep 28 2020 08:21:57 GMT+0800 (China Standard Time)

That suggests that this has nothing to with CVXR directly. Your installation does not run a function from the gmp package which CVXR uses. So there is something wrong with your R installation or the packages you have installed.

Eduardo Horta · Answer 11 · Mon Sep 28 2020 09:53:53 GMT+0800 (China Standard Time)

I'm closing this issue since the problem seems to be stemming from the gmp package: indeed, writing

library(gmp)
gmp::as.bigq(3)

yields the output

 *** caught illegal operation ***
address 0x10d320f80, cause 'illegal opcode'

Traceback:
 1: gmp::as.bigq(3)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace

I reported the gmp bug on https://forgemia.inra.fr/sylvain.jasson/gmp/-/issues/1

JohannesFriedrich · Answer 12 · Wed Dec 16 2020 03:28:39 GMT+0800 (China Standard Time)

I had the same problems and the solution was to clone the repository from the R-package gmphere:
https://forgemia.inra.fr/sylvain.jasson/gmp.git

and then I built the package by myself and all problems with macOS and the gmp package were gone.

Eduardo Horta · Answer 13 · Wed Dec 16 2020 04:45:23 GMT+0800 (China Standard Time)

@JohannesFriedrich Could you please detail how you proceeded?

JohannesFriedrich · Answer 14 · Wed Dec 16 2020 05:04:16 GMT+0800 (China Standard Time)

I think you have git, a c compiler and gmp (eg. via brew) installed on your mac.

Remove your current R-package gmp via remove.packages("gmp")
Go to the command line
write git clone https://forgemia.inra.fr/sylvain.jasson/gmp.gitand clone the repository
write R CMD build gmp to build the package
write R CMD install gmp to install the package

See the command line if any error occurs. If not, go to R and test, e.g. gmp::factorize(12)

Eduardo Horta · Answer 15 · Wed Dec 16 2020 05:51:17 GMT+0800 (China Standard Time)

@JohannesFriedrich I was able to build but not install:

$ R CMD build gmp
* checking for file ‘gmp/DESCRIPTION’ ... OK
* preparing ‘gmp’:
* checking DESCRIPTION meta-information ... OK
* cleaning src
* running ‘cleanup’
* checking for LF line-endings in source and make files and shell scripts
* checking for empty or unneeded directories
* looking to see if a ‘data/datalist’ file should be added
* building ‘gmp_0.6-1.tar.gz’

$ R CMD install gmp
* installing to library ‘/Library/Frameworks/R.framework/Versions/4.0/Resources/library’
* installing *source* package ‘gmp’ ...
** using staged installation
checking for gcc... gcc
checking whether the C compiler works... yes
checking for C compiler default output file name... a.out
checking for suffix of executables... 
checking whether we are cross compiling... no
checking for suffix of object files... o
checking whether we are using the GNU C compiler... yes
checking whether gcc accepts -g... yes
checking for gcc option to accept ISO C89... none needed
checking for __gmpz_ui_sub in -lgmp... no
configure: error: GNU MP not found, or not 4.1.4 or up, see http://gmplib.org
ERROR: configuration failed for package ‘gmp’
* removing ‘/Library/Frameworks/R.framework/Versions/4.0/Resources/library/gmp’

Any insights?

JohannesFriedrich · Answer 16 · Wed Dec 16 2020 05:53:06 GMT+0800 (China Standard Time)

The easiest way: use brew install gmp from your command line (if you have brew installed!).