On Mon, 22 Apr 2002, Jacky Lam wrote:

> Actually, I have both version...But I think that the patch I sent
> should be faster in terms of number of syscalls while the overhead of
> realloc in both case should be more or less the same.
>
> Through I am not sure, the difference between two different ways
> should be small and the speed overhead of this part is not so important to
> the overall performance of libcurl. Do you think so?

Thinking about it a bit more, I think you're right. Doing multiple realloc()s
to enlarge the buffer will mean more badness than this.

I tried out a Solaris version that seems to work fine and I have applied your
patch with my minor edits now.

Thanks a lot!

-- 
    Daniel Stenberg -- curl groks URLs -- http://curl.haxx.se/