Dear Alexander! I applied the patch to version 3.5.15, and it works fine! Thanks a lot! I'll look out for your changes-notifications to see which future LFTP-Release will include the patch.
By the way: after having reported this problem, I found out that LFTP _does_ stop in the described scenario, but after a timeout of about ten minutes. So some other timeout seems to have applied?! By the way (2): the mirror-command doesn't stop after the first fatal error (see debug output below), although --max-errors=1 is set. That's not critical, of course -- even if the network connection is restored after the first fatal error and no other fatal error occurs, the LFTP return code indicates that mirroring was not successful. Kind regards Michael ---> CWD /hhla/releases/Release_vc_client_2.5/common.external.runtime-23.0/libs/ swiftmq <--- 250 CWD successful. "/hhla/releases/Release_vc_client_2.5/common.external.r untime-23.0/libs/swiftmq" is current directory. ---> TYPE A **** Timeout - reconnecting ---- Closing data socket ---- Closing control socket ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket mirror: Fatal error: max-retries exceeded Mirroring directory `common.jms-14.3' Making directory `common.jms-14.3' ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket ---- Connecting to <host> port 21 **** Timeout - reconnecting ---- Closing control socket mirror: Fatal error: max-retries exceeded >>> "Alexander V. Lukyanov" <[EMAIL PROTECTED]> 08.11.2007 08:46 >>> On Thu, Nov 01, 2007 at 10:23:12AM +0100, Michael Reidt wrote: > I tested the previously reported case (concerning 3.5.13) of a connection > loss during directory upload (mirror) with 3.5.15 and still have the same > trouble: the timeout detection doesn't seem to work in certain situations. > Should I test this scenario with 3.6 also, or will the newest release act the > same way? Here is a patch to fix the problem. -- Alexander.
