On Tue, 15 Sep 2015, David Chapman wrote:

>> If you're using libcurl as a win32 DLL, you MUST use the
>> CURLOPT_WRITEFUNCTION if you set this option or you will experience
>> crashes.

>> Due to how ownership of that FILE pointer is handled.

> I don't think the wording here is at all clear. Are you saying that when
> cURL is used as a Win32 DLL, fwrite is not a valid CURLOPT_WRITEFUNCTION
> even if passed in explicitly?

Hm, yes I guess that is the logical conclusion and it is a good question. I'm
not sure, maybe our way of using it makes it different? The problem has
something to do with the FILE * being owned by the application and it can't
pass it on to the DLL to be accessed from there, but the flow isn't 100% clear
to me.

> Or are you saying that use of NULL as a CURLOPT_WRITEFUNCTION is not valid
> when cURL is used as a Win32 DLL, and that a pointer to fwrite must be
> passed in (or else the call should be skipped)?

The regular way we've phrased it, is that you need to implement a write
callback. I'm not sure what happens if you just put fopen there.

All this is just info I've gathered and learned from other libcurl users, I am
myself not an experienced wanderer in the world of Windows memory setups and
concepts.

> In either case, the documentation in the CURLOPT_WRITEFUNCTION page should
> have the same warning. You don't want pages to be inconsistent with each
> other.

Agreed!

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html