Requesting feedback and testers for Windows platform

[kyleb]

I would like to request feedback from mrgsolve users on Windows platforms. Our regular use is exclusively Mac/unix, but we do some testing on Windows.  Unfortunately, it's not as comprehensive as I would like.  Any input from users so far would be very valuable for improving mrgsolve on that platform.

There is an issue in 0.5.002 with compiling the model when there are spaces in the project path that I suspect some have experienced.  There is a wiki entry here: https://github.com/metrumresearchgroup/mrgsolve/wiki/Windows-Issues including workaround.  (This isn't strictly a Windows issue, but maybe more likely on that platform).

Better yet: if you are interested / able to test a development version in Windows, please post here and I'll send you instructions.  This development version fixes the spaces issue, some other smaller bugs, and several new features.

Thanks,

Kyle

[Kyle Baron]

Is anyone having issues installing either Rcpp or mrgsolve (or anything other R package) from source on a Windows machine?

If you've been successful, what is your PATH environment variable?

I have been successfully installing and testing mrgsolve and all dependencies from source with latest Rtools and this PATH:

> Sys.getenv("PATH")
[1] "C:\\Program Files\\R\\R-3.2.3\\bin\\x64;c:\\Rtools\\bin;c:\\Rtools\\gcc-4.6.3\\bin;c:\\Rtools\\bin;c:\\Rtools\\MinGW64\\bin;C:\\Windows\\system32;C:\\Windows;C:\\Windows\\System32\\Wbem;C:\\Windows\\System32\\WindowsPowerShell\\v1.0\\;C:\\Program Files\\R\\R-2.13.1\\bin;C:\\Program Files\\TortoiseSVN\\bin;C:\\cygwin\\bin;"

[Kyle Baron]

Thanks for posting, Amy.  It looks like the same sort of stuff I've got in my PATH.  I think I had to set the paths myself a long time ago but couldn't recall the details of what I actually did.  

For others struggling with the compiler on Windows: I think getting the path right is really important.  Merely installing Rtools is not enough.

It it not possible to read this too carefully:

https://cran.r-project.org/doc/manuals/R-admin.html#The-Windows-toolset

To quote:

Your PATH may include . first, then the bin directories of the tools, the compiler toolchain and LaTeX. Do not use filepaths containing spaces: you can always use the short forms (found by dir /x at the Windows command line). Network shares (with paths starting \\) are not supported.

 

For example for a 32-bit build, all on one line,
PATH=c:\Rtools\bin;c:\Rtools\gcc-4.6.3\bin;c:\MiKTeX\miktex\bin;
c:\R\R-3.2\bin\i386;c:\windows;c:\windows\system32

 

It is essential that the directory containing the command line tools comes first or second in the path: there are typically like-named tools⁶³ in other directories, and they will not work. The ordering of the other directories is less important, but if in doubt, use the order above.
Our toolset contains copies of Cygwin DLLs that may conflict with other ones on your system if both are in the path at once. The normal recommendation is to delete the older ones; however, at one time we found our tools did not work with a newer version of the Cygwin DLLs, so it may be safest not to have any other version of the Cygwin DLLs in your path.

There is also some nice help on the Rstan site:

https://github.com/stan-dev/rstan/wiki/Install-Rtools-for-Windows

https://github.com/stan-dev/rstan/issues/141

I'll keep posting new resources here or new bits of information that might help future users.  And maybe a help page of our own (but it would / should look exactly like the Rstan help).

[Devin Pastoor]

Hi Kyle,

I have found that the latest (non-frozen version of Rtools) seems to cause all sorts of problems for compilation, however have had very good success (installing my PKPDmisc package with some Cpp functions through Rcpp) across a couple hundred windows computers in various corporate/personal settings with no issues if you use the latest frozen release (so one release prior to most recent) that matches your R version. This can also be autodetected through the installr() package. Key is to do a full restart after as well and I haven't had a single person not work after those 2 steps.

I haven't tried mrgsolve on windows yet but can give it a shot later to see if the same behavior holds.

Devin

--

MetrumRG Website: http://www.metrumrg.com/opensourcetools.html
---
You received this message because you are subscribed to the Google Groups "mrgsolve" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mrgsolve+u...@metrumrg.com.
To post to this group, send email to mrgs...@metrumrg.com.
Visit this group at https://groups.google.com/a/metrumrg.com/group/mrgsolve/.

[Kyle Baron]

Devin - 

Thanks a lot for the insight.  So that means (for now):

https://cran.r-project.org/bin/windows/Rtools/

Rtools32.exe or Rtools31.exe

I most recently tested under:

R version 3.2.3 (2015-12-10)

Platform: x86_64-w64-mingw32/x64 (64-bit)

Running under: Windows Server 2008 R2 x64 (build 7601) Service Pack 1

--

Kyle Baron

Metrum Research Group

860-735-7043, Ext. 202

ky...@metrumrg.com

[Vijay Ivaturi]

During the R courses when I helped install PKPDmisc, my experience is that it is best to use the Rtools31.exe, but mainly via the installr() package. The installr() function does a much better job of setting the paths on any sort of computer as opposed to the regular installation. 

Vijay

[Devin Pastoor]

I can officially confirm that with a fresh R 3.2.2 install with Rtools32.exe mrgsolve installs fine on windows10.

On Thu, Feb 25, 2016 at 10:39 PM, Devin Pastoor <devin....@umaryland.edu> wrote:

correct, so I'd use Rtools32.exe, even though Rtools33.exe (non-frozen) is supposedly for R 3.2.x the non-frozen versions have always given headaches. I'll check on my windows 10 desktop tomorrow for mrgsolve, though this pattern has held for PKPDmisc's cpp dependencies/ dplyr/etc from github

ps - another little nugget - if a package on github doesn't have any fortran/c/c++ code you don't need rtools to install from github (handy for data-only or small scale packages if rtools install is ever a bottleneck)

On Thu, Feb 25, 2016 at 10:34 PM, Kyle Baron <ky...@metrumrg.com> wrote:

Devin - 

Thanks a lot for the insight.  So that means (for now):

[Kyle Baron]

installr looks great ... going to start using that. 

Is this what you call?

install.packages("installr")
install.Rtools()

When it sets paths, does it just modify .Renviron or set things outside of R too? 

[Devin Pastoor]

after installing you can just

installr::installr() and it will pop open a gui that you can choose Rtools

or do

installr::install.installr() for programmatic install

As to how it works, you can see https://github.com/talgalili/installr/blob/master/R/install.R#L501 that it does the same method I use (taking the most recent frozen version) for the install.

More details at

https://github.com/talgalili/installr

Note: installR doesn't change/set any paths - the system path doesn't matter I don't think, as the key is that R knows where to look for the rtools, which you can check via devtools::find_rtools() 

On Thu, Feb 25, 2016 at 10:49 PM, Kyle Baron <ky...@metrumrg.com> wrote:

installr looks great ... going to start using that. 

Is this what you call?

install.packages("installr")
install.Rtools()

When it sets paths, does it just modify .Renviron or set things outside of R too? 

On Thu, Feb 25, 2016 at 9:39 PM, Vijay Ivaturi <ivija...@gmail.com> wrote:

During the R courses when I helped install PKPDmisc, my experience is that it is best to use the Rtools31.exe, but mainly via the installr() package. The installr() function does a much better job of setting the paths on any sort of computer as opposed to the regular installation. 

Vijay

On Thu, Feb 25, 2016 at 10:34 PM Kyle Baron <ky...@metrumrg.com> wrote:

Devin - 

Thanks a lot for the insight.  So that means (for now):

httpsUMB Campus policy prohibits embedded exe linksor Rtools31.exe

[Buchwald, Amy]

When installing Rtools, check the “Edit the system PATH” box.

 

 

From: Kyle Baron [mailto:ky...@metrumrg.com]
Sent: Thursday, February 25, 2016 6:20 PM
To: Buchwald, Amy; mrgsolve
Subject: Re: [mrgsolve] Re: Requesting feedback and testers for Windows platform

 

Thanks for posting, Amy.  It looks like the same sort of stuff I've got in my PATH.  I think I had to set the paths myself a long time ago but couldn't recall the details of what I actually did.  

 

For others struggling with the compiler on Windows: I think getting the path right is really important.  Merely installing Rtools is not enough.

 

It it not possible to read this too carefully:

https://cran.r-project.org/doc/manuals/R-admin.html#The-Windows-toolset

 

 

To quote:

Your PATH may include . first, then the bin directories of the tools, the compiler toolchain and LaTeX. Do not use filepaths containing spaces: you can always use the short forms (found by dir /x at the Windows command line). Network shares (with paths starting \\) are not supported.

 

For example for a 32-bit build, all on one line,
PATH=c:\Rtools\bin;c:\Rtools\gcc-4.6.3\bin;c:\MiKTeX\miktex\bin;
c:\R\R-3.2\bin\i386;c:\windows;c:\windows\system32

 

It is essential that the directory containing the command line tools comes first or second in the path: there are typically like-named tools⁶³ in other directories, and they will not work. The ordering of the other directories is less important, but if in doubt, use the order above.
Our toolset contains copies of Cygwin DLLs that may conflict with other ones on your system if both are in the path at once. The normal recommendation is to delete the older ones; however, at one time we found our tools did not work with a newer version of the Cygwin DLLs, so it may be safest not to have any other version of the Cygwin DLLs in your path.

 

 

 

There is also some nice help on the Rstan site:

https://github.com/stan-dev/rstan/wiki/Install-Rtools-for-Windows

https://github.com/stan-dev/rstan/issues/141

 

I'll keep posting new resources here or new bits of information that might help future users.  And maybe a help page of our own (but it would / should look exactly like the Rstan help).

 

 

 

 

 

 

On Thu, Feb 25, 2016 at 7:01 AM, Buchwald, Amy <ABuc...@noven.com> wrote:

Hi Kyle,

 

I was able to successfully install it.

[path was redacted]

 

Sincerely,

                     Amy

--

MetrumRG Website: http://www.metrumrg.com/opensourcetools.html
---
You received this message because you are subscribed to the Google Groups "mrgsolve" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mrgsolve+u...@metrumrg.com.
To post to this group, send email to mrgs...@metrumrg.com.
Visit this group at https://groups.google.com/a/metrumrg.com/group/mrgsolve/.

[Kyle Baron]

I suspect that some folks may be having problems with Rtools installation because they don't have proper access to set the system PATH on their computer.  

Looking at the installr project:

https://github.com/talgalili/installr/blob/master/R/install.R#L437

it doesn't seem to be doing anything special to set up the path in the current or future R session (like through .Renviron).  So if the user is being blocked from changing the PATH, this might still be an issue. 

I put some simple instructions up here to set your PATH within R:

https://github.com/metrumresearchgroup/mrgsolve/wiki/Windows-issues#set-a-path-for-your-r-session

These are pretty simple (minded) but any user should be able to do this for their R session and seemed to work for one person who is having issues.  

But ... please do remember to manually check your path in the R session (Sys.getenv("PATH")) and / or try some simple compilation exercise like this: https://gist.github.com/kylebmetrum/d5b9865cdc49e2ea630a or this https://github.com/stan-dev/rstan/wiki/Install-Rtools-for-Windows If these tests don't work it means you don't have things set up properly and you need to go back to the destructions and read carefully, following each and every step ... no matter how small it may seem.  If you haven't checked and are having problems, please check now.

[mct941]

That's right.  Installr didn't do anything for me since the system PATH was locked down on my corporate laptop.  It would bring me to the same error about not being able to change the registry when executing RTools, even when I checked the "Edit system PATH" box.  Changing the PATH under the Windows User Accounts > Tasks > Change my Environment Variables didn't work as a workaround since it place the RTools path in the middle of the System PATH.  Below is a workaround that seemed to work for me using R v3.2.0 and RTools v3.2.0.  Hope this helps.

Sys.getenv("PATH") # no RTools added here

PATH <- Sys.getenv("PATH")

PATH.1 <- "C:\\Rtools\\bin\\"

PATH.2 <- "C:\\Rtools\\gcc-4.6.3\\bin\\"

Sys.setenv("PATH"=paste(PATH.1,PATH.2,PATH,sep=";"))

Sys.getenv("PATH") # RTools path added here

install.packages("Rcpp",type = "source",repos="http://cran.us.r-project.org")

install.packages("RcppArmadillo",type = "source",repos="http://cran.us.r-project.org")

install.packages("BH",type = "source",repos="http://cran.us.r-project.org")

install.packages("dplyr",type = "source",repos="http://cran.us.r-project.org")

install.packages("~/mrgsolve_0.5.001.tar.gz", repos = NULL, type = "source")

[Vijay Ivaturi]

I can confirm that the setup described in the previous email works on a windows laptop. Thank you for sharing. I tested on two different systems with Rtools3.2 and it works great.

Install Rtools3.2 

Sys.getenv("PATH") # no RTools added here

PATH <- Sys.getenv("PATH")

PATH.1 <- "C:\\Rtools\\bin\\"

PATH.2 <- "C:\\Rtools\\gcc-4.6.3\\bin\\"

Sys.setenv("PATH"=paste(PATH.1,PATH.2,PATH,sep=";"))

Sys.getenv("PATH") # RTools path added here

install.packages("Rcpp",type = "source",repos="http://cran.us.r-project.org")

install.packages("RcppArmadillo",type = "source",repos="http://cran.us.r-project.org")

install.packages("BH",type = "source",repos="http://cran.us.r-project.org")

install.packages("dplyr",type = "source",repos="http://cran.us.r-project.org")

install.packages("~/mrgsolve_0.5.001.tar.gz", repos = NULL, type = "source")

[Anna Olsén]

I'm using Windows 10 on a small computer I bought just for PKPD analysies... :)
Started learning all this stuff in February, so, new to it all, but so far, everythhing mostly works.

It is indeed taking a bit of time getting to grips with it though.

So, I'm testing it all out on a Windows10, so I'll be able to do some testing later on, once I delivered my first results... :)

Anna

[Kyle Baron]

Anna - 

Please let me know if we can help with anything.  

GitHub is the preferred location these days: https://github.com/metrumresearchgroup/mrgsolve/issues

Thanks,

Kyle