There have been several threads in the forums and several bugs opened and closed about emerge sync failing with I/O errors/timeouts. Yet despite all this the problem still keeps recurring. It can be fine for weeks at a time, then it comes back again for anything up to a couple of days where emerge sync will not complete.
The problem (at least as it affects me) seems to be that it firstly downloads the names (and I presume checksums/MD5/etc) of >100,000 files which form the portage tree. The local system then checks to see if any new files need downloading or existing files updated. If while it is doing this, a period of time (typically 180s) elapses when no changes are requested (ie there is no traffic on the open TCP link to the mirror) the mirror closes the TCP connection and the local system shows a timeout error. My experience is that setting RSYNC_TIMEOUT (as documented in /etc/make.conf.example) in /etc/make.conf has no effect and does not affect the time which the connection has to be idle before it times out. Can nothing be done (even if just making the mirrors respect RSYNC_TIMEOUT) to address this intermittent but long standing problem? -- [email protected] mailing list
