[webkit-dev] Proposal: Avoiding fastMalloc(0) in WebKit

[Eric Seidel]

I have posted a patch to (optionally) ASSERT whenever fastMalloc is
called with 0 size:
https://bugs.webkit.org/show_bug.cgi?id=55097

After fixing https://bugs.webkit.org/show_bug.cgi?id=55091 for a large
performance win in the HTML parser (by avoiding a single fastMalloc(0)
call) I now believe that every call to fastMalloc(0) is a programming
error in our source.  It should always be faster to branch in the
caller to avoid fastMalloc(0) than to call fastMalloc(0).

I am announcing this in case anyone has opinions or comments, as once
ports start turning this on, you may hit this assertion with your
code.

I welcome your commentary on the bug!

-eric