Closed quickbeam123 closed 3 months ago
Thanks for creating the issue! I have a question though. So far I thought the udpr=off
variant is the correct one, which means we are incomplete rather than unsound?
You are right. I changed the issue's title.
Does #564 fix this issue and the duplicate?
The problem is fixed on the examples provided in both issues, so I would consider this solved.
On: 00082minimized.txt we have master with
--decode ott+2_1:1_updr=on:foolp=on_2
giving saturation, while with--decode ott+2_1:1_updr=off:foolp=on_2
we find a proof.@mezpusz is already after this one with https://github.com/vprover/vampire/compare/master...avoid-symbol-elimination-for-smtlib2-defined-symbols so, mostly, this issue tracks that we don't forget.
However, in addition to / instead of marking the relevant symbols as protected, it would be nice for the specially SMT-parsed function definitions to be turned back into normal formulas AT THE BEGINNING of the preprocessing pipeline under
-fnrw off
while their special treatment later on in the pipeline would only kick in for-fnrw on
. Note that it might be necessary to mark both the actually parsed symbol as well as its "proxy" (named with the sFN /sPN prefix).