I think that emerge should whine at me when fetching fails. This behaviour should be default, even in quiet mode. A parameter could control it. Backstory: For some seemingly random packages, emerge spent up to three hours doing seemingly nothing. I have know figured it out and fixed the problem. The problem was that my mirror has apparently died on me, and so fetching was not exactly working. If emerge had whined at me, I would have spotted this at once. (Arguably I would have spotted it at once if I were running tail -f /var/log/emerge-fetch.log as well, but then again I'm lazy.)
We could use different parameters to control behavior for different modes of failure. For example, checksum failures are probably the most common mode of failure that you would be interested in. If it's an issue with the mirror timing out, I don't think there's much we can do there, since generally a human has to tail -f /var/log/emerge-fetch.log in order to detect that kind of thing.
We can have an argument for when to timeout, and also when to warn. So you could set WARN_ME_WHEN to 120, and if nothing happens it 120 secs, emerge says "mirror foo is taking suspiciously long...", and then sett FAIL_WHEN to 240 and automatically die when 240 secs have gone.
(In reply to Alexander Berntsen from comment #2) > We can have an argument for when to timeout, and also when to warn. So you > could set WARN_ME_WHEN to 120, and if nothing happens it 120 secs, emerge > says "mirror foo is taking suspiciously long...", and then sett FAIL_WHEN to > 240 and automatically die when 240 secs have gone. But, we don't want it to bail out early if the download is large relative to the connection speed. I wonder if there are some wget options for this. Since portage simply executes FETCHCOMMAND, it's not really possible for portage to monitor progress, unless we replace FETCHCOMMAND with python library calls.
wget has -T, which is the network timeout in seconds. It also has --dns-timout, --connect-timout and --read-timeout. Is it not possible though to check the downloaded file at WARN_WHEN and FAIL_WHEN? i.e. something has been downloaded, at all ? well all right then : WARN/FAIL.
(In reply to Alexander Berntsen from comment #4) > Is it not possible though to check the downloaded file at WARN_WHEN and > FAIL_WHEN? i.e. something has been downloaded, at all ? well all right then > : WARN/FAIL. Actually, yes, it would be possible for portage to monitor the file size.
There's some example code in pym/portage/tests/ebuild/test_ipc_daemon.py the would serve as a good starting point for monitoring the file size. You would use SpawnProcess to spawn FETCHCOMMAND, and replace EbuildIpcDaemon with a different class that monitors the file size.
