On Wed, 20 Mar 2002, Jean-Philippe Barrette-LaPierre wrote:

> ok I will resume all the stuff:
>
> thereis :
> CURLOPT_DEBUGDATA
> CURLOPT_STDERR
> CURLOPT_DEBUGFUNCTION

Yeps, and for completeness, this is also affected:

  CURLOPT_VERBOSE

> CURLOPT_DEBUGDATA is set by default on STDERR (witch is made when we set it
> to NULL, and of course at the beginning)

I don't think we need to do that. The DEBUGDATA can be made independent of
STDERR. Let it be NULL by default.

> CURLOPT_STDERR is set by default on stderr (witch is made when we set it to
> NULL, and of course at the beginning)

> CURLOPT_DEBUGFUNCTION is set by default on my_debug() (witch is made when
> we set it to NULL, and of course at the beginning), BUT it will reset the
> CURLOPT_DEBUGDATA to CURLOPT_STDERR. The cause is that the callback change,
> so the datat need it too (we presume it will be different type of data).

I think that the default built-in DEBUGFUNCTION my_debug() could use whatever
STDERR is set to to output its information. It doesn't need to use the
DEBUGDATA at all.

I want the CURL handle passed to the callback, remember? And using that
handle, the built-in callback has every mean to extract all sorts of random
info it wants!

The my_debug() function will also probably ignore the most verbose debug
levels (data in/out) too, as they're not output at all today.

> CURLOPT_DEBUGFUNCTION is used when the verbose option is on.

Yeps.

> So then, the debug output will be set to stderr on default, but can be set
> to others things.

Right. Preserving the old functionality for existing applications is very
important.

> all this stuff is okay, or I am loosing my mind?

Hah, this stuff seems okay, but that doesn't necessariliy mean that you
aren't losing your mind anyway! ;-P

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