You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, as we found out in #4914 the TR_S390BinaryAnalyser and TR_S390BinaryCommutativeAnalyser work in different ways w.r.t. decrementing node counts. The former seems to decrement always, while the latter does not. We need to standardize on which way we should lean.
Since the analyzer ends up evaluating the children my preference would be for the analyzer to be responsible for decrementing the children as well.
The text was updated successfully, but these errors were encountered:
fjeremic
changed the title
Determine whether analyzers should decrememnt children node counts or now
Determine whether analyzers should decrememnt children node counts or not
Mar 9, 2020
0xdaryl
changed the title
Determine whether analyzers should decrememnt children node counts or not
Determine whether analyzers should decrement children node counts or not
Mar 10, 2020
The convention that nodes are decremented where they are evaluated is historically how the backends were architected to work (although the rules have been bent in a number of places and not necessarily for the good). I agree the analyzers should decrement the ref counts on the children it evaluates.
Currently, as we found out in #4914 the
TR_S390BinaryAnalyser
andTR_S390BinaryCommutativeAnalyser
work in different ways w.r.t. decrementing node counts. The former seems to decrement always, while the latter does not. We need to standardize on which way we should lean.Since the analyzer ends up evaluating the children my preference would be for the analyzer to be responsible for decrementing the children as well.
The text was updated successfully, but these errors were encountered: