0% coverage reported on Windows

[maelle]

I am using the development version of covr which solved this issue I had with a previous version.

When running covr::package_coverage() for this package with tests I get 0% coverage which puzzles me given 1) codecov results 2) the fact that tests actually run on my laptop via devtools::test().

I am not sure which info to include in this issue. :-)

[wlandau-lilly]

@maelle I could not reproduce this issue with 3c24869, 32-bit R-devel (2017-09-23 r73342), and Windows 7. For me, coverage is still 100% for drake. Do you also get 0% coverage with other packages?

> sessionInfo()
R Under development (unstable) (2017-09-23 r73342)
Platform: i386-w64-mingw32/i386 (32-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

Matrix products: default

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

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

other attached packages:
[1] covr_3.0.1

loaded via a namespace (and not attached):
[1] compiler_3.5.0 lazyeval_0.2.1 magrittr_1.5   rex_1.1.2     
[5] tools_3.5.0    withr_2.1.0    rstudioapi_0.7 yaml_2.1.14   
[9] crayon_1.3.4 

[maelle]

Good question, yes I get it with other packages, I have just tried that.

I get this warning

Warning messages:
1: running command 'C:/PROGRA~1/R/R-34~1.0PA/bin/x64/R --vanilla CMD config CC' had status 1 
2: running command 'C:/PROGRA~1/R/R-34~1.0PA/bin/x64/R --vanilla CMD config CC' had status 1 

> sessionInfo()
R version 3.4.0 Patched (2017-05-10 r72669)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

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

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

loaded via a namespace (and not attached):
[1] compiler_3.4.0 lazyeval_0.2.0 magrittr_1.5   rex_1.1.2      tools_3.4.0   
[6] withr_2.1.0    rstudioapi_0.6 crayon_1.3.4   covr_3.0.1 

will update lazyeval and rstudioapi.

[maelle]

Update: updating these two packages didn't make a difference.

[wlandau-lilly]

I tried to meet you half-way: 64-bit R 3.4.0. After update.packages(), I got 100% coverage for this package (quicker to test).

From your warning messages, I see your R installations are in C:/Program Files/R/. For different issues, the space in that file path was the culprit, so now I keep all my R versions in C:/R/. Not sure if it would help in this case, but if covr is spawning child processes, it might.

[maelle]

Oh interesting! I'm not going to change the R location right now, but it's good to know.

[wlandau-lilly]

I am glad you did not jump at the suggestion because C:/Program Files/ does not seem to be the problem. I tried installing R 3.4.2 there, and I still got the expected full coverage.

[jimhester]

@maelle can you try running system("C:/PROGRA~1/R/R-34~1.0PA/bin/x64/R --vanilla CMD config CC") directly? I think this is an issue with interpreting short names on the OS, not sure what the solution is, we have similar issue in devtools (r-lib/devtools#885) with some possible user side fixes.

[maelle]

`

system("C:/PROGRA1/R/R-341.0PA/bin/x64/R --vanilla CMD config CC")
"sh" no se reconoce como un comando interno o externo,
programa o archivo por lotes ejecutable.
Warning message:
running command 'C:/PROGRA1/R/R-341.0PA/bin/x64/R --vanilla CMD config CC' had status 1 `

So sh is not recognized, whatever that means?

[wlandau-lilly]

What if you updated Rtools? It's a shot in the dark, but it seems like the easiest next thing to try.

[randomsven]

... just to avoid "special moments" - I also had a similar problem to you (if I read correctly): running covr on a collection of R scripts on Windows (red herring) produced 0 coverage. The code I was trying to check test coverage for was being migrated into a package, but was originally a set of scripts - tests had to leverage the source function to drag sets of functions in, which overwrites the code that has been instrumented by covr - getting rid of any source statements and building the package solved the problem. ... not saying this is your issue, but it had me stumped for a little while :)

[robne1982]

If it helps, I have the exact same issue as @maelle . Not sure there's much I can add, but I'd be happy to try out any suggestions anyone has?

UPDATE - I tried installing the latest version of Rtools too.

My version is

R version 3.4.2 (2017-09-28)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

[maelle]

@wlandau-lilly Thank you, I just ran installr::install.Rtools() that told me I have the relevant version installed. 🤔

@randomsven Thanks but in my case the exact same packages gets a different coverage on other systems. 😉

@rnobleeddy Thank you, I'm happy to not be alone in this position and I hope we'll soon get it sorted out!

[tonyelhabr]

I'm having the same issue.

[thomaslaber]

Same here.

[idefer]

I don't know if it may be related, but I was getting the same issue while running file_coverage(), using it for one source file and one test file. It turned out, I had rm(list = ls()) at the top of the test file to clean up the environment before running the tests. Removing that line fixed the issue.

[hartmast]

I have the same issue on a Mac, both with the CRAN and the Github versions; I tried running package_coverage() on several different packages.

EDIT: Apologies, I found out that the error was entirely on my side - I had .gitignored the test files; after changing this, everything works like a charm.

[q23qweliuhan]

I got similar issue here, just create a new package with the hello.R, and created test using use_test().
After pointing to the correct Rtools directory, I don't get the warnings on the CMD config anymore. But I still get this 0% coverage. Apparently they are not related.

> devtools::test()
Loading test
Testing test
√ | OK F W S | Context
√ |  1       | test-example.R

== Results =====================================================================
Duration: 0.2 s

OK:       1
Failed:   0
Warnings: 0
Skipped:  0
> package_coverage()
test Coverage: 0.00%
R/hello.R: 0.00%

[jimhester]

So I think these errors tend to occur because covr expects that R and RTools are installed on your PATH. If they are not you need to use something like devtools::setup_rtools() to put it on the PATH prior to running covr.

[jimhester]

This should now be fixed, although you may still get similar installation errors if you try to run covr on a package needing compilation without Rtools being on the path.