Set total verdict and total similarity more precisely on base of automatic marks assotiation
There are the following proposals:
- Calculate total similarity as a maximum.
- Prevent incompatible marks setting and total similarity recalculation if there is at least one manually confirmed mark exists for this error trace.
- Do not consider 0% marks calculating total verdict to prevent useless incompatible marks label.
#2 Updated by Evgeny Novikov 2 months ago
I think that if we will agree to implement #9412 in another way, this issue will gone. BTW, it demonstrates other problems when "external" marks can lead to unplesant statistics.
One more observation, I already suggested somewhere to calculate a total similarity as a maximum, but then I thought more and decided that the current way is better. Indeed, it can help to detect mark duplicates when some mark patterns catch error traces but similarities are not 100%. Sometimes, even fewer similarities are okay and one can delete extra marks with 100% similarity. Besides, you can find "bad" marks with <100% similarity and remove them since new marks match error traces better.