On Mon, 26 Sep 2011 15:54 Daniel Stenberg <daniel_at_haxx.se> wrote:

> On Mon, 26 Sep 2011, Marcin Adamski wrote:
>
> > After TFTP download is started, I drop connection with server by
> > disconnecting Ethernet cable. In CURLOPT_PROGRESSFUNCTION I check the time
> > of last change of NowDownloaded parameter value of this callback. If more
> > than 5seconds elapsed since last change, I return 1 in this callback,
> > because I assume that connection is dropped.
>
> ...
>
> > This is a workaround for no support for CURLOPT_LOW_SPEED_TIME and
> > CURLOPT_LOW_SPEED_LIMIT options in cURL's TFTP implementation.
>
> I'd say that lack of support is an oversight that we should fix!
>
> > The problem is that transfer is not finished immediately and
> > CURLOPT_PROGRESSFUNCTION is called again and again until transfers actually
> > ends after 1891.87 s.
>
> So there's a bug that makes the transfer not end correctly even though it
> should. Any chance you can help us write up a test case for the curl test
> suite that repeats this problem?
>
>

If there is a documentation about writing test cases for the curl test suite, I could try doing it in a free time.

Could you give me some estimate when this problem can will be fixed? I need it for now and I woulnd't like to make some dirty hacks for this problem. I'm thinking also about fixing it in libcurl by myself, but I'm not sure if it's a good idea, because it could take me too much time.

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2011-09-26