nsCRT::HashCode(const PRUnichar*) has shown up on a bunch of profiles as a significant chunk of total time (0.2% on a page load profile, 0.4% on a new window profile). I'd like to make it much simpler and move the current implementation to HashCodeAsUTF8. I don't know of anyone who depends on the behavior that it gives the same hash code as nsCRT::HashCode(const char*) does for the equivalent UTF8 string. scc, do you know of any users who depend on that behavior? There's a patch to do this within attach 65095 on bug 118933.

er, attachment 65095 [details] [diff] [review]

To clarify, the patch to review is the second half (last 2 files, nsCRT.*) of attachment 65095 [details] [diff] [review].

sr=jst

r=jag

Fix checked in 2002-02-16 11:50 PST.