[bugfix] Fix bug in unloading of conflicting modules

[teojgo]

• Fix bug which led to accumulation of module unloading.

• Use raw string to avoid warnings.

Fixes #59

[codecov-io]

Codecov Report

Merging #405 into master will increase coverage by <.01%.
The diff coverage is 80%.

@@            Coverage Diff             @@
##           master     #405      +/-   ##
==========================================
+ Coverage   91.23%   91.23%   +<.01%     
==========================================
  Files          70       70              
  Lines        8581     8582       +1     
==========================================
+ Hits         7829     7830       +1     
  Misses        752      752

Impacted Files	Coverage Δ
reframe/core/environments.py	89.93% <80%> (+0.06%)	⬆️
unittests/test_policies.py	99.54% <0%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b5cad8f...f338b26. Read the comment docs.

[vkarak]

Can you explain what was exactly the bug and how it was triggered precisely?

[teojgo]

Imagine that module A conflicts with module C and module B conflicts with D. If we wanted to load modules A, B, looping through them would do the following:

1. self._conflicted = []
2. after checking the conflicts of A: self._conflicted = ['C']
3. Therefore before loading A, module C would be unloaded
4. after checking the conflicts of B: self._conflicted = ['C', 'D']
5. Since self._conflicted belongs to the instance, before loading B, modules C, D would be unloaded although C was unloaded in 3.

[vkarak]

@teojgo Good, I get it now.

[vkarak]

The loop variable should be simply m.

[teojgo]

its a nested loop and the outer one already uses m

[vkarak]

Move this assignment after the loop. Placed here, it kind of lets you expect that self._conflicted will be used just right after.

[vkarak]

I'd name this simply conflicted.