Bug 927705 made a change to pldhash that is worth propagating to HashTable. HashTable.h currently has a 16M capacity limit. This is currently a hard limit due to the use of 8-bit fixed point arithmetic (which fails above 16M) to compute min/max loads. The 8-bitness is unnecessary complexity and should be removed. We could consider increasing the max capacity limit as well (pldhash now has 64M). If we do, the following static assertion JS_STATIC_ASSERT((sMaxCapacity * sizeof(Entry)) <= UINT32_MAX); might need to be reconsidered; if it's too onerous, we could change to a dynamic overflow check, which is what pldhash uses.

Note to self: HashTable has two unit tests in js/src/jsapi-tests/: testHashTable.cpp and testHashTableInit.cpp, and I don't understand why the testHashInitTooHuge() in the latter fails -- it seems like a bigger init number should be allowed. Maybe it's a limit of the 8-bit fixed point arithmetic.

Another idea worth stealing from pldhash is bug 933074 -- allowing reasonable overloading (past 75% full) if the table cannot be grown.