The Static Analysis tool Coverity detected that |lhs| is not initialised on all paths. For example: >> if (handler.isUnparenthesizedDestructuringPattern(lhs)) { >> if (kind != PNK_ASSIGN) { >> error(JSMSG_BAD_DESTRUCT_ASS); >> return null(); >> } I think it wouldn't hurt having it initialised right after it's declaration: >>Node lhs = null();

Comment on attachment 8825794 [details] Bug 1330296 - initialise |lhs| with default null value. Unsurprisingly, Coverity's over-warning here -- |lhs| is only uninitialized if |maybeAsyncArrow| is true, and that's true only if, at the end of the |if (maybeAsyncArrow)| block, the current token is TOK_ARROW. And that causes the subsequent switch to always go into a TOK_ARROW case that never uses |lhs| and returns at the end of the block that the case labels. This is sort of a fix, but more it's a placation. However, my patches in bug 1319416 stand a good chance of fixing this a better way -- have the same basic control flow with |if (tt != TOK_ARROW) fail;| at the end of |if (maybeAsyncArrow)|, but then switch on |tt| directly in the subsequent switch. Let's get those patches reviewed/landed, then see if this warning shows up still.