Hi Olivier, I think that whenever someone is doing something following the instructions given in 1804.10017 and the code is doing something different than what is explained in that reference not only a warning but a stop and a request to do set acknowledged_v3.1_syntax True --global should appear. This indeed happens already with generate p p > t t~ QCD=1 QED=1 [/*QED*/] but it does not with generate p p > t t~ QCD=1 QED=1 [/*QCD*/] I do not understand why there is a difference in the two cases. In the first one it generates amplitudes contributing to LO_2 and add corrections order alpha to the diagrams only, leading to an incomplete and IR divergent NLO_3 (p p > t t~ QCD=0 QED=2 [/*QCD*/] is missing). The situation is probably even worse because the only amplitude available is    a g > t t~. In the second case it generates again amplitudes contributing t LO_2 and add corrections order alpha_s, leading to an incomplete and IR divergent NLO_2 (p p > t t~ QCD=2 QED=0 [/*QED*/] is missing). So the real question is: why the reaction of the code to an old-style syntax is different in the two previous cases? Both are not doing what is explained in 1804.10017. If instead I do  p p > t t~ [QED] or p p > t t~ QCD=99 QED=99 [QCD QED] I obtain what I expect from 1804.10017. So indeed no warning is necessary, no stop is necessary. Concerning the part in which I mentioned Marco, I was asking a confirmation to what you say. If indeed there is no difference with the old syntax, warning should not be there. Cheers Davide On 10.05.21 13:18, Olivier Mattelaer wrote: > Hi, > > Yes as I said in our last meeting (and in the merge request), I have weaken a bit the scope in which we crash and provide only a warning for a series of case where the syntax was already possible in 2.x and were the paper 1804.10017 is not an issue per say. Do you think that a warning is not enough? > >> Marco: should also QCD=99 AND QED=99 [QCD] or QCD=99 AND QED=99 [QED] raise an error? > In this case they are no ambiguity and therefore I would say that even a warning is not needed. > This is for me the same case as if the user does not provide anything. > I strongly believe that we should not put warning if we know that syntaxes between 2.x, 3.x and 3.1.x are the same. > Useless warning are the best way for people to ignore them. > > > Cheers, > > Olivier > > >> On 10 May 2021, at 12:54, davide.pagani.85