mrgsolve error compiling models

[tmss1]

I've been using mrgsolve for a while now. I've recently updated R to v4.0.2 with rtools40 and the latest mrgsolve on CRAN, and started seeing this error when compiling with this code that ran without any problems before that update. Does anyone know what's wrong?

Building m23 ... Error in file(con, "r") : cannot open the connection

[kylebaron]

Can you compile anything? Or just m23?

[tmss1]

I have a few models to compile, m23 is the frist one. I tried skipping m23 as a test but the others have the same error.

[kylebaron]

Ok; I'd look at the Rtools40 installation; were you able to install a package from source? See the example below:

[tmss1]

Yes, I followed the examples below from the rtools site and no problem installing from the source.

[kylebaron]

Sounds good. Happy to keep working with you on this. Let me know if there are any other leads to investigate.

[tmss1]

Thanks Kyle. Since that's the only error message I got, I'm not sure what else I can look into at this point.

[kylebaron]

You could share more information:

• what does the output look like when you compiled the package from source?
• can you debug mread and let us know at what point you get the error (?debugonce)?
• did you blow away everything from your R 3.x install? Could your system be looking at / for old files?
• what is the sessionInfo()?
• what do you get with pkgbuild::check_build_tools()?

Etc ... there must be more information that could be shared that might suggest what is happening on your system.

[kylebaron]

• Is this a network drive?
• Are you working with a network drive at all?
• Are there spaces in the path to R?
• Does this work: tmp <- tempfile(); cat(letters,file = tmp); readLines(tmp) (can you write to tempdir())?'
• This: pkgbuild::check_compiler(debug = TRUE)?

[kylebaron]

@tmscs Any luck with any of this?

[tmss1]

Thanks for the tips Kyle. Apology for my delayed response. I've now managed to test your suggestions as below.

what does the output look like when you compiled the package from source?

This is the output when I installed the testing json package on the rtool40 website:

:/rtools40/mingw64/bin/gcc -shared -s -static-libgcc -o jsonlite.dll tmp.def base64.o collapse_array.o collapse_object.o collapse_pretty.o escape_chars.o integer64_to_na.o is_datelist.o is_recordlist.o is_scalarlist.o modp_numtoa.o null_to_na.o num_to_char.o parse.o prettify.o push_parser.o r-base64.o register.o row_collapse.o transpose_list.o validate.o -Lyajl -lstatyajl -LC:/PROGRA~1/R/R-4.0.2/bin/x64 -lR
installing to C:/Program Files/R/R-4.0.2/library/00LOCK-jsonlite/00new/jsonlite/libs/x64
** R
** inst
** byte-compile and prepare package for lazy loading
** help
*** installing help indices
converting help for package 'jsonlite'
finding HTML links ... done
base64 html
flatten html
fromJSON html
prettify html
rbind_pages html
read_json html
serializeJSON html
stream_in html
unbox html
validate html
** building package indices
** installing vignettes
** testing if installed package can be loaded from temporary location
*** arch - i386
*** arch - x64
** testing if installed package can be loaded from final location
*** arch - i386
*** arch - x64
** testing if installed package keeps a record of temporary installation path
** DONE (jsonlite)

The downloaded source packages are in
‘C:\Users\más\AppData\Local\Temp\RtmpMdwROc\downloaded_packages’

can you debug mread and let us know at what point you get the error (?debugonce)?

The is the output when I tried debugonce. I've omitted the model code since it's long and probably not the issue.

debugging in: mcode_cache("m23", s23code)
debug: {
mread_cache(model, project, code = code, ...)
}
Browse[2]> model
[1] "m23"
Browse[2]> project
[1] "C:\Users\Tomás\AppData\Local\Temp\RtmpmGKSPx"
Browse[2]> code
[1] "\n$PARAM @Annotated......."
Browse[2]>
debug: mread_cache(model, project, code = code, ...)
Browse[2]>
Building m23 ... Error in file(con, "r") : cannot open the connection

did you blow away everything from your R 3.x install? Could your system be looking at / for old files?

I think so.

what is the sessionInfo()?

I've checked this before and couldn't detect any irregularities:

R version 4.0.2 (2020-06-22)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19041)

Matrix products: default

locale:
[1] LC_COLLATE=English_Australia.1252
[2] LC_CTYPE=English_Australia.1252
[3] LC_MONETARY=English_Australia.1252
[4] LC_NUMERIC=C
[5] LC_TIME=English_Australia.1252

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

other attached packages:
[1] ggpubr_0.4.0 scales_1.1.1 ggplot2_3.3.2 mrgsolve_0.10.4
[5] magrittr_1.5 tidyr_1.1.1 dplyr_1.0.1 plyr_1.8.6

loaded via a namespace (and not attached):
[1] zip_2.1.0 Rcpp_1.0.5
[3] cellranger_1.1.0 pillar_1.4.6
[5] compiler_4.0.2 forcats_0.5.0
[7] tools_4.0.2 lifecycle_0.2.0
[9] tibble_3.0.3 gtable_0.3.0
[11] pkgconfig_2.0.3 rlang_0.4.7
[13] openxlsx_4.1.5 curl_4.3
[15] haven_2.3.1 rio_0.5.16
[17] RcppArmadillo_0.9.900.2.0 withr_2.2.0
[19] generics_0.0.2 vctrs_0.3.2
[21] hms_0.5.3 grid_4.0.2
[23] tidyselect_1.1.0 glue_1.4.1
[25] data.table_1.13.0 R6_2.4.1
[27] rstatix_0.6.0 readxl_1.3.1
[29] foreign_0.8-80 carData_3.0-4
[31] purrr_0.3.4 car_3.0-9
[33] backports_1.1.8 ellipsis_0.3.1
[35] abind_1.4-5 colorspace_1.4-1
[37] ggsignif_0.6.0 stringi_1.4.6
[39] munsell_0.5.0 broom_0.7.0
[41] crayon_1.3.4

what do you get with pkgbuild::check_build_tools()?

pkgbuild::check_build_tools()
Your system is ready to build packages!

Is this a network drive?

No.

Are you working with a network drive at all?

No.

Are there spaces in the path to R?

Only the space in Programe Files as here: "C:\Program Files\R\R-4.0.2\bin\x64"
This has not caused problem before though.

Does this work: tmp <- tempfile(); cat(letters,file = tmp); readLines(tmp) (can you write to tempdir())?'

This is the output. The accent in my name is the one causing the encoding problem in the path below. I've been using mrgsolve on this computer with earlier verions of R before so not sure if it indicates any problem:

tmp <- tempfile(); cat(letters,file = tmp); readLines(tmp)
[1] "a b c d e f g h i j k l m n o p q r s t u v w x y z"
Warning message:
In readLines(tmp) :
incomplete final line found on 'C:\Users\más\AppData\Local\Temp\RtmpOOBqqi\file4841dc859e5'

This: pkgbuild::check_compiler(debug = TRUE)?

This is the output:

pkgbuild::check_compiler(debug = TRUE)
Trying to compile a simple C file
Running "C:/PROGRA~1/R/R-4.0.2/bin/x64/Rcmd.exe" SHLIB foo.c
"C:/rtools40/mingw64/bin/"gcc -I"C:/PROGRA~1/R/R-4.0.2/include" -DNDEBUG -O2 -Wall -std=gnu99 -mfpmath=sse -msse2 -mstackrealign -c foo.c -o foo.o
C:/rtools40/mingw64/bin/gcc -shared -s -static-libgcc -o foo.dll tmp.def foo.o -LC:/PROGRA~1/R/R-4.0.2/bin/x64 -lR
[1] TRUE

[tmss1]

I've been using another machine for now while trying to solve this. I've checked through the outputs above for quite a few times now and the output from debugonce seems to be clearer now and I've updated that above. I just can't figure out what's issue though. Did I miss anything obvious above?

[kylebaron]

Thanks, @tmscs ; I'm sorry I missed your output from 2 weeks ago; I'll look at it this morning after a couple of meetings and get back to you. It does look like most of the stuff is in place and working.

[kylebaron]

I wonder if this is the culprit: "C:\Users\Tomás\AppData\Local\Temp\RtmpmGKSPx".

Can you try this:
Create a directory: C:/project

then

mod <- mcode_cache("m23", s23code, project = "C:/project")

I'd like to know what happens there.

Additionally:

Create a directory: C:/build

then

mod <- mcode_cache("m23", s23code, project = "C:/project", soloc = "C:/build")

[kylebaron]

Rdatatable/data.table#3078

[kylebaron]

mod <- mcode_cache("m23", s23code, project = enc2native(tempdir()))

tmp <- enc2native(tempdir())
mod <- mcode_cache("m23", s23code, project = tmp, soloc = tmp)

[tmss1]

Thanks so much Kyle, it works now! The minimum change for things to work is by specifying "soloc":

mod <- mcode_cache("m23", s23code, soloc = "C:/build")

With the above, it works even without specifying the project folder. It did not work when I only specified the project folder without "soloc". So it looks like "soloc" here is the issue.

This latter code here also did not work with the same error:

tmp <- enc2native(tempdir())
mod <- mcode_cache("m23", s23code, project = tmp, soloc = tmp)

May I ask if you know what is the reason for this error, so I can avoid this in the future?

[kylebaron]

Thanks for finding the minimal change. That's really helpful.

soloc is the place where mrgsolve compiles the model code; it needs to write out some files and compile. It might also need to unload or remove a model dll file. It does this inside tempdir(), but inside a couple of subfolders specific to the model name and platform you are running on.

When you use mread_cache, it just writes the model file (what is in s23code) to that directory (here tempdir())

So I wondered if this was some problem with the path to tempdir() but it looks like that isn't it (since tempdir() works fine for project).

I guess it could be some permissions issue, but all of those directories for soloc should be getting created by you so there shouldn't be a reason why you couldn't access them.

Short term fix

Just keep using this. You can put the following

options(mrgsolve.soloc = "C:/build")

at the top of your script or in .Rprofile in your HOME directory.

But obviously I'd like to figure out why this is happening and fix the root cause so you don't always have to do this. I can throw out some ideas but I'm afraid it will be tedious. Would you be willing to so a screen share some time and try some things together? I think we could make a lot of ground pretty quick. If you'd be ok with that, could you email me at the address in my github profile?

Thanks,
Kyle

[tmss1]

Yes, we can look into this when you have a moment. Feel free to let me know how I can help.