R SIG Debian - May 2020 - Sometimes commands do not terminate after upgrading to R 4.0 and Ubuntu 20.04

I have upgraded R (from 3.6 to 4.0) and RStudio (from 1.1 to 1.2.5) a few
days ago, and Ubuntu from 18.04 to 20.04 yesterday.

Since then, R sometimes never terminates when executing certain commands:
ivreg (from package AER), summary (of a logit regression) and logitmfx
(from package mfx). Sometimes these commands run fine, but most of the time
I have to kill the process because R won't terminate the execution, even
when pressing the red Stop button in RStudio or doing Ctrl + C when using R
in the terminal.

When I tried example('AER'), it worked fine. Then I re-installed the
package AER. It threw 10 warnings of type In readLines(file, skipNul TRUE) : 
cannot open compressed file
'/usr/lib/R/site-library/[package]/DESCRIPTION', probable reason 'No
such
file or directory' where [package] is abind, colorspace, dichromat... (but
not AER).

Since then example('AER') throws a warning: no help found for ?AER?.

I've removed and reinstalled R 4.0: it didn't help. I've downgraded
to R
3.6: it didn't help. I've tried both answering yes or No when asked to
install the packages in a personal library: none works. So I think the
problem comes from Ubuntu 20.04, and is related to this post:
https://stat.ethz.ch/pipermail/r-sig-debian/2020-April/003166.html. When I
reinstall packages, I have a bunch of warnings related to Fortran 2018 that
no longer support certain commands.

I hope someone can help.

-- 
Adrien Fabre

?cole d'?conomie de Paris/Paris School of Economics ? Universit? Paris 1
(R4-47)

Page personnelle/Home page <http://sites.google.com/view/adrien-fabre>
(+33/0)6.10.37.90.51

	[[alternative HTML version deleted]]

Am Mittwoch, 13. Mai 2020, 14:19:07 CEST schrieb Adrien
FABRE:
> I have upgraded R (from 3.6 to 4.0) and RStudio (from 1.1 to 1.2.5) a few
> days ago, and Ubuntu from 18.04 to 20.04 yesterday.
> 
> Since then, R sometimes never terminates when executing certain commands:
...
> install the packages in a personal library: none works. So I think the
> problem comes from Ubuntu 20.04, and is related to this post:
> https://stat.ethz.ch/pipermail/r-sig-debian/2020-April/003166.html.
...

Did you try 

sudo update-alternatives --config libblas.so.3-x86_64-linux-gnu

Johannes

	[[alternative HTML version deleted]]

Salut Adrien,

On 13 May 2020 at 14:19, Adrien FABRE wrote:
| I have upgraded R (from 3.6 to 4.0) and RStudio (from 1.1 to 1.2.5) a few
| days ago, and Ubuntu from 18.04 to 20.04 yesterday.
| 
| Since then, R sometimes never terminates when executing certain commands:
| ivreg (from package AER), summary (of a logit regression) and logitmfx
| (from package mfx). Sometimes these commands run fine, but most of the time
| I have to kill the process because R won't terminate the execution, even
| when pressing the red Stop button in RStudio or doing Ctrl + C when using R
| in the terminal.
| 
| When I tried example('AER'), it worked fine. Then I re-installed the
| package AER. It threw 10 warnings of type In readLines(file, skipNul | TRUE) :
cannot open compressed file
| '/usr/lib/R/site-library/[package]/DESCRIPTION', probable reason
'No such
| file or directory' where [package] is abind, colorspace, dichromat... (but
| not AER).
| 
| Since then example('AER') throws a warning: no help found for ?AER?.
| 
| I've removed and reinstalled R 4.0: it didn't help. I've
downgraded to R
| 3.6: it didn't help. I've tried both answering yes or No when asked to
| install the packages in a personal library: none works. So I think the
| problem comes from Ubuntu 20.04, and is related to this post:
| https://stat.ethz.ch/pipermail/r-sig-debian/2020-April/003166.html. When I
| reinstall packages, I have a bunch of warnings related to Fortran 2018 that
| no longer support certain commands.
| 
| I hope someone can help.

Yes! See an email thread here started on the (Debian "internal" but of
course
public) list debian-science by me on April 30 and carried into May
so you need both

  https://lists.debian.org/debian-science/2020/04/msg00081.html
  https://lists.debian.org/debian-science/2020/05/msg00000.html

as the thread spanse two months.

It appears to be a bad OpenMP and and OpenBLAS interaction you can (for now)
avoid) by replacing the 'pthread' variant of OpenBLAS with the OpenMP
version
(see the thread for details).  Doing

  sudo apt install  libopenblas-openmp-dev
  sudo apt remove libopenblas-pthread-dev

should do -- make sure you do the second step too to ensure that the pthread
variant is remove as it would rank higher. Check via `R -s -e
'sessionInfo()'`
and look closely at the path printed for blas+lapack. Without pthread you
should be good.

Seb, CC'ed, and who is I have that right ma actually pretty close to you via
the PSE, is actually looking for someone with the proper hardware to create a
proper bug report. I was able to reproduce this once on my systems (both
Ubuntu and Debian testing) but was unable later. Some hardware dependence
still likely.

So in short: yes, we're aware, there is a workaround, but we could use help
in pinning it down. And sorry for the trouble. Apparently an old bug reborn.

Amicalement,  Dirk

PS Am a AMSE grad myself though it had a different name at the time ;-)
 
| -- 
| Adrien Fabre
| 
| ?cole d'?conomie de Paris/Paris School of Economics ? Universit? Paris 1
| (R4-47)
| 
| Page personnelle/Home page <http://sites.google.com/view/adrien-fabre>
| (+33/0)6.10.37.90.51
| 
| 	[[alternative HTML version deleted]]
| 
| _______________________________________________
| R-SIG-Debian mailing list
| R-SIG-Debian at r-project.org
| https://stat.ethz.ch/mailman/listinfo/r-sig-debian

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org

Thank you very much Dirk!



Le mer. 13 mai 2020 ? 14:59, Dirk Eddelbuettel <edd at debian.org> a ?crit
:
>
> Salut Adrien,
>
> It appears to be a bad OpenMP and and OpenBLAS interaction you can (for
> now)
> avoid) by replacing the 'pthread' variant of OpenBLAS with the
OpenMP
> version
> (see the thread for details).  Doing
>
>   sudo apt install  libopenblas-openmp-dev
>   sudo apt remove libopenblas-pthread-dev
>
This works! Thank's a lot!

should do -- make sure you do the second step too to ensure that the
pthread
> variant is remove as it would rank higher. Check via `R -s -e
> 'sessionInfo()'`
> and look closely at the path printed for blas+lapack. Without pthread you
> should be good.
>
The command yields:
BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.9.0
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/liblapack.so.3
so pthread still seems there in LAPACK, but I think the bug solved
nonetheless.

> Seb, CC'ed, and who is I have that right ma actually pretty close to
you
> via
> the PSE, is actually looking for someone with the proper hardware to
> create a
> proper bug report. I was able to reproduce this once on my systems (both
> Ubuntu and Debian testing) but was unable later. Some hardware dependence
> still likely.
>
What a small world! Happy to help with a bug report: how can I do that?
>
> So in short: yes, we're aware, there is a workaround, but we could use
help
> in pinning it down. And sorry for the trouble. Apparently an old bug
> reborn.
>
> Amicalement,  Dirk
>
> PS Am a AMSE grad myself though it had a different name at the time ;-)
>
Great, happy to see that economists can also code!
>
> | --
> | Adrien Fabre
> |
> | ?cole d'?conomie de Paris/Paris School of Economics ? Universit?
Paris 1
> | (R4-47)
> |
> | Page personnelle/Home page
<http://sites.google.com/view/adrien-fabre>
> | (+33/0)6.10.37.90.51
> |
> |       [[alternative HTML version deleted]]
> |
> | _______________________________________________
> | R-SIG-Debian mailing list
> | R-SIG-Debian at r-project.org
> | https://stat.ethz.ch/mailman/listinfo/r-sig-debian
>
> --
> http://dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org
>

-- 
Adrien Fabre

?cole d'?conomie de Paris/Paris School of Economics ? Universit? Paris 1
(R4-47)

Page personnelle/Home page <http://sites.google.com/view/adrien-fabre>
(+33/0)6.10.37.90.51

	[[alternative HTML version deleted]]