Bourbaki marks this one as harder than the rest.
Let R be a root system, C a chamber. Put B(C)={α1,…,αl},
R+(C)={α1,…,αl,αl+1,…,αs}
(the αi being pairwise distinct), and α=α1+⋯+αs.
a) For i=1,2,…,s, put εi=±1. Let α∗=ε1α1+⋯+εsαs. If (α∗∣αi)>0 for i=1,…,l, then α∗=α and εi=1 for all i. (Let γ (resp. δ) be the sum of the αi for which εi=1 (resp. −1). Then, (γ∣αi)−(δ∣αi)>0, (γ∣αi)+(δ∣αi)=(αi∣αi), so
2(γ∣αi)>(αi∣αi);
since 2(γ∣αi)/(αi∣αi)∈Z, (γ∣αi)≥(αi∣αi)=(α∣αi), so γ−α∈C, γ≥α, γ=α and δ=0.
b) For i=1,2,…,s, let εi=±1. The set of εiαi is the set of roots >0 relative to a chamber if and only if α∗=∑iεiαi belongs to a chamber. (To show that the condition is sufficient, let μi=±1 be such that (α∗∣μiαi)>0. There exists an element w∈W(R) that transforms the set of μiαi into the set of αi, and hence α∗ into α∗∗=∑iεiμiασ(i), where σ∈Ss. Applying a), show that α∗∗=α, so μiεi=1.)