fix: solution.Error() returns untyped nil when SAT

[joelanford]

Fixes #139

[codecov]

Codecov Report

Merging #140 (1b37541) into main (4c15c99) will increase coverage by 0.13%.
The diff coverage is 100.00%.

❗ Current head 1b37541 differs from pull request most recent head f58565c. Consider uploading reports for the commit f58565c to get more accurate results

@@            Coverage Diff             @@
##             main     #140      +/-   ##
==========================================
+ Coverage   61.90%   62.03%   +0.13%     
==========================================
  Files          16       16              
  Lines         567      569       +2     
==========================================
+ Hits          351      353       +2     
  Misses        198      198              
  Partials       18       18              

Files Changed	Coverage Δ
pkg/deppy/solver/solver.go	84.21% <100.00%> (+0.87%)	⬆️

[m1kola]

I think we should just use error interface for s.err instead a specific type. This way we avoid having this strange if and still can avoid nil not being nil issue.

diff --git a/pkg/deppy/solver/solver.go b/pkg/deppy/solver/solver.go
index d60036f..f8758f5 100644
--- a/pkg/deppy/solver/solver.go
+++ b/pkg/deppy/solver/solver.go
@@ -16,7 +16,7 @@ import (
 // A successful execution of the solver can still end in an error when no solution can
 // be found.
 type Solution struct {
-       err       deppy.NotSatisfiable
+       err       error
        selection map[deppy.Identifier]deppy.Variable
        variables []deppy.Variable
 }

[joelanford]

Done! Wasn't sure if that change would have knock-on effects. Looks like it compiles and tests pass though, so I assume all good.