corecfg: handle unknown keys that are explicitly set

[zyga]

When updateKeyValueStream is processing a stream with a key that is not
known but is explicitly listed in new configuration it would needlessly
indicate that the file needs writing.

The test that was checking this was flaky because the writes happened
so quickly we were never given a chance to observe the failure
(usually). Running the test repeatedly on my SSD makes it fail very
frequently tough.

Signed-off-by: Zygmunt Krynicki zygmunt.krynicki@canonical.com

[codecov-io]

Codecov Report

Merging #3746 into master will increase coverage by 0.11%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #3746      +/-   ##
==========================================
+ Coverage   75.72%   75.84%   +0.11%     
==========================================
  Files         398      399       +1     
  Lines       34235    34530     +295     
==========================================
+ Hits        25926    26190     +264     
- Misses       6459     6477      +18     
- Partials     1850     1863      +13

Impacted Files	Coverage Δ
corecfg/utils.go	93.75% <100%> (+0.72%)	⬆️
cmd/snap-repair/runner.go	82.96% <0%> (-3.85%)	⬇️
interfaces/sorting.go	98.71% <0%> (-1.29%)	⬇️
asserts/asserts.go	90.92% <0%> (ø)	⬆️
interfaces/builtin/browser_support.go	76.31% <0%> (ø)	⬆️
interfaces/builtin/optical_drive.go	100% <0%> (ø)	⬆️
interfaces/builtin/pulseaudio.go	37.03% <0%> (ø)	⬆️
interfaces/builtin/udisks2.go	100% <0%> (ø)	⬆️
interfaces/builtin/alsa.go	100% <0%> (ø)	⬆️
asserts/store_asserts.go	98.52% <0%> (ø)
... and 6 more

---

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 eac3d5d...1758807. Read the comment docs.

[mvo5]

Thanks for the PR! I looked into this some more and I think we want a slightly different semantic here. The idea behind "allConfig" is really "supportedConfigOptions", i.e. it contains the map of the things updateKeyValueStream understands. I think we want to error if "newConfig" contains things that are not n "allConfig". I prepare a branch for this now (with some better tests) and will also rename "allConfig" into "supportedConfigKeys".