On Tue, 29 May 2007, Yang Tse wrote:

> I know that this is like comparing apples with oranges. Does the speed
> improvement the intermediate var introduces outweight the compiler warning
> it also introduces?

Not really, we should probably revert my patch in that regard.

The reason I changed it was that the previous version had the off-by-one due
to the unconditional writing to poll_fds[ix].events field even when ix was
increased to the poll_nfds value.

It was just after that fix I realized it could just as well break out of the
loop when completed, and then the previous change actually wasn't required
anymore.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html