Join GitHub today
runtime: audit mapfast routines for indirectvalue support #21547
On reflection, the most recent CLs I mailed suggest that mapdelete_fast* might not be handing indirect value maps correctly. I'm out of computer time for today and maybe tomorrow as well. Filing this as a reminder, and because these were newly introduced in 1.9, and if they're not well-handled, it could cause memory corruption. Should double check all routines, really.
@josharian indirect values always use the slow map: