Empty solution error description

[pawelru]

r$> pkgdepends::new_pkg_installation_proposal(c("cran/dm@1.0.0", "cran/dplyr@1.0.9")) %>% {.$solve(); .$show_solution()}
Error: 
! Could not solve package dependencies:

Type .Last.error to see the more details.

This is for lpSolve_5.6.17. If I do upgrade to the newest version 5.6.18 I even got a red error:

r$> pkgdepends::new_pkg_installation_proposal(c("cran/dm@1.0.0", "cran/dplyr@1.0.9")) %>% {.$solve(); .$show_solution()}
Error in .C("lpslink", direction = as.integer(direction), x.count = as.integer(x.count),  : 
  Incorrect number of arguments (29), expecting 28 for 'lpslink'

[gaborcsardi]

The first one seems like a bug indeed. I cannot reproduce the second one, maybe you didn't restart R after the installation?

[gaborcsardi]

If you use a new library, then you'll see the real reason:

❯ p <- pkgdepends::new_pkg_installation_proposal(c("cran/dm@1.0.0", "cran/dplyr@1.0.9"), config = list(library = tempfile()))
ℹ Creating library directory:  /var/folders/ph/fpcmzfd16rgbbk8mxvy9m2_h0000gn/T//RtmpMsCIiD/fileaa583ad9ccc9

❯ p$solve()
❯ p$get_solution()
<pkg_solution>
+ result: FAILED
+ refs:
  - cran/dm@1.0.0
  - cran/dplyr@1.0.9
+ constraints (165):
  - select dm exactly once
  - select dplyr exactly once
  - select DBI at most once
  - select Matrix at most once
  - select R6 at most once
  - select backports at most once
  - select cachem at most once
  - select cli at most once
  - select cpp11 at most once
  - select fansi at most once
  ...
x failures:
* cran/dm@1.0.0: Can't install dependency tidyr (>= 1.0.0)
* tidyr: Can't install dependency dplyr (>= 1.0.10)
* dplyr: Conflicts with cran/dplyr@1.0.9

[pawelru]

Thank you. After session restart my red error is gone. Adding config arg also solved blank report issue but I guess this ticket is still valid as we want to have it without that arg specified.

[pawelru]

I actually have one more use case of a blank failure report and I got it with tempfile() as you suggested.

r$> temp_desc <- desc::desc("!new")

r$> temp_desc$set_dep("dm", "Imports")

r$> temp_desc$set_dep("dplyr", "Imports")

r$> temp_desc$set_remotes(c("cran/dm@1.0.0", "cran/dplyr@1.0.9"))

r$> temp_desc
Package: {{ Package }}
Title: {{ Title }}
Version: 1.0.0
Authors@R (parsed):
    * Jo Doe <jodoe@dom.ain> [aut, cre]
Maintainer: {{ Maintainer }}
Description: {{ Description }}
License: {{ License }}
URL: {{ URL }}
BugReports: {{ BugReports }}
Imports:
    dm,
    dplyr
Remotes:
    cran/dm@1.0.0,
    cran/dplyr@1.0.9
Encoding: UTF-8

r$> temp_desc_path <- tempfile()

r$> temp_desc$write(temp_desc_path)

r$> x <- pkgdepends::new_pkg_deps(temp_desc_path, config = list(library = tempfile()))

r$> x$solve()

r$> x$get_solution()
<pkg_solution>
+ result: FAILED
+ refs:
  - local::/var/folders/m1/hrz0h_ls7gz57rc80tnj41t80000gp/T//Rtmp0cwFMJ/fileaffd42c31262
+ constraints (166):
  - select {{ Package }} exactly once
  - select dm at most once
  - select dplyr at most once
  - select DBI at most once
  - select Matrix at most once
  - select R6 at most once
  - select backports at most once
  - select cachem at most once
  - select cli at most once
  - select cpp11 at most once
  ...
x failures:

Above works perfectly fine without failures - you would just need to comment-out one of the imports.

Do you have any idea how to make it work?

[gaborcsardi]

These are the same versions as above, no? So the error is the same, tidyr needs dplyr 1.0.10:

* cran/dm@1.0.0: Can't install dependency tidyr (>= 1.0.0)
* tidyr: Can't install dependency dplyr (>= 1.0.10)
* dplyr: Conflicts with cran/dplyr@1.0.9

[pawelru]

Yes. Those are the same references and after your last comment I know the reason. I just reused it to give you another example of a erroneous behavior - this time using also library = tempfile() as you suggested.

[pawelru]

Hi @gaborcsardi is there any plans to attack this bad guy? It's sort of relevant for my project as it's hard to interpret empty error reports.

[gaborcsardi]

Reprex:

temp_desc <- desc::desc("!new")
temp_desc$set_dep("dm", "Imports")
temp_desc$set_dep("dplyr", "Imports")
temp_desc$set_remotes(c("cran/dm@1.0.0", "cran/dplyr@1.0.9"))
temp_desc_path <- tempfile()
temp_desc$write(temp_desc_path)
x <- pkgdepends::new_pkg_deps(temp_desc_path, config = list(library = tempfile()))
x$solve()
x$get_solution()

[gaborcsardi]

With b37d08e this now reports:

x failures:
* local::/var/folders/ph/fpcmzfd16rgbbk8mxvy9m2_h0000gn/T//RtmpVhwlCY/file26cf53f5773c: dependency conflict

[pawelru]

Thank you @gaborcsardi. I do confirm that using @main I can see now "dependency conflict" error message instead of the empty one. I agree with your commit message that hopefully this can be improved soon as I think that the package itself is definitely able to point to the right root-cause of the conflict. Therefore I would like to keep this ticket open.

For anyone else looking at this ticket. I managed to find out sort of workaround solution to put a vector of refs read from package description as opposed to single reference to the package itself. Using reprex above:

temp_desc <- desc::desc("!new")
temp_desc$set_dep("dm", "Imports")
temp_desc$set_dep("dplyr", "Imports")
temp_desc$set_remotes(c("cran/dm@1.0.0", "cran/dplyr@1.0.9"))
temp_desc_path <- tempfile()
temp_desc$write(temp_desc_path)
x <- pkgdepends::new_pkg_deps(temp_desc_path, config = list(library = tempfile()))
x$solve()
x$get_solution()

x2 <- pkgdepends::new_pkg_deps(temp_desc$get_remotes(), config = list(library = tempfile()))
x2$solve()
x2$get_solution()