On Tue, 9 Dec 2014, Patrick Monnerat wrote:

> I noticed another potential problem in Curl_client_write() with the pause
> feature:

In fact, pausing alone might very well be reason enough for us explicitly not
allowing code to use custom buffers and sizes. Or we need to complicate
Curl_client_write() more in order to handle multiple calls with additional
data to hold during pause...

> - Let's start with an unpaused input transfer.

Yes, let's take one step at a time.

> - When resuming from the pause, the pushed data is translated again, because
> flushed by the Curl_client_write() procedure :-(

Good thing then that the translation thing isn't that commonly used. We
obviously need to stick to a single way to do it though.

> I'll look to fix this problem together with the long size, because these are
> quite related... But it might take a bit longer than a thought. Any comment
> ?

I would like to hear from Bill and Steve if they think it would be doable to
change the SMB code to use the existing 'buffer' and 'uploadbuffer' memory
areas instead of the custom ones? It would solve these issues and avoid more
quirks and funny work-arounds.

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