Studies of user sentiment on social networks like Twitter have formed a steadily growing research area. But there is still lack of knowledge on whether the discussion clusters tagged by emotionally opposite hashtags differ in sentiment distribution, both in terms of difference between hashtags and between user types, e.g. non-influencers and influential accounts. We look at two hashtags that marked the discussion on the Charlie Hebdo massacre of 2015, namely #jesuischarlie and #jenesuispascharlie. As sentiment analysis studies for the French language are rare, we elaborate our own approach to sentiment vocabulary. We apply human coding and machine learning to correct the automated sentiment assessment. Then we apply the enhanced knowledge on sentiment to both discussion segments and compare the configuration of the resulting sentiment-based nebulae in overall and francophone-only discussions. Also, we define influencers for both discussions and compare whether ordinary and institutional users differ by sentiment. We have three notable findings. First, negativity structures #jenesuispascharie more than #jesuischarlie. Second, while francophones communicate cross-sentiment inside the francophone talk, their negativity tends to cast impact upon cluster formation inside general discussions. Third, influencers in both cases tend to be more negative than positive, but institutional users bear neutral and positive sentiment more than ordinary people.