Yes, I can confirm that when reverting that commit, the timeout is triggered by curl. As far as I can tell, with valgrind I couldn't see any memory leaks from libssh2 in this case, but I just ran it a couple of times so I wouldn't fully trust my small test to conclude that the memory leak is not there anymore.
Are you planning on reverting that commit in libcurl? On Fri, May 22, 2020 at 2:37 PM Daniel Stenberg <dan...@haxx.se> wrote: > On Wed, 20 May 2020, Estanislau Augé-Pujadas via curl-library wrote: > > > I have been using curl to upload files with sftp, and I have encountered > an > > unexpected behaviour related to timeouts and disconnects. When setting > > CURLOPT_TIMEOUT_MS to some value, and the network disconnects during a > file > > upload, libcurl does not timeout the operation. > > > > I found a commit on Github that indicates that this behaviour is > actually > > intentional to address a problem that could leak to memory leaks for > libssh2 > > > https://github.com/curl/curl/commit/f31760e63b4e9ef1eb25f8f211390f8239388515 > > Yeah, when I'm looking at it now it seems like the wrong fix (that I made > and > pushed). If libssh2 does indeed leak that memory, the fix should be done > in > libssh2. We should not just ignore a timeout request from a user. > > If you revert that change, does the timeout work properly? Does libssh2 > (still) leak memory if you do? > > -- > > / daniel.haxx.se | Commercial curl support up to 24x7 is available! > | Private help, bug fixes, support, ports, new features > | https://www.wolfssl.com/contact/
------------------------------------------------------------------- Unsubscribe: https://cool.haxx.se/list/listinfo/curl-library Etiquette: https://curl.haxx.se/mail/etiquette.html
