On 2017-11-23 18:13, db wrote: > On 23 Nov 2017, at 17:06, Ryan Schmidt <ryandes...@macports.org> wrote: >> ftp connections fail in many circumstances, due to either the use or non-use >> of passive mode, or for other reasons. I'm certainly tired of attempting to >> debug them, which is why it's great when we can remove the use of ftp from a >> portfile and replace it with http(s). > > Before the change, it should have timed out in 30", not in 30'. Why didn't it?
I still have no idea what these units are, please communicate clearly. Always provide debug logs as that would have helped to identify FTP much earlier. I can confirm that the previous Portfile version that used ftp:// for livecheck also hangs longer than 30 seconds for me. It hangs the same way with curl(1) on command line, after EPSV. The timeout we configure is only for the initial connection, but the transfer itself may take as long as it takes. I am not sure why the configured speed limit does not cause the transfer to be aborted after a while, though... We could provide an additional flag for our 'curl fetch' API in Pextlib to set an overall timeout with CURLOPT_TIMEOUT that is only used for livecheck. However, I do not think it is worth to invest effort into that. Hosts or protocols that fail for some users are a bad option value anyway. It is better not to use FTP at all. Rainer