On 04.10.2006, at 07:24, Mark Duling wrote:
Can I get a vote on whether the snippet below from bug 10226 represents a feature request or a bug? Being the bug Nazi is a thankless job. :)
[The port database is sometimes left in an inconsistent state when the port command is aborted] IMHO: It's a bug, plain and simple. It has hit me before. It's not even documented anywhere that I can see. If you think we cannot fix it at the moment, we should at least trap all signals to show that it's our "intention" not to have port interrupted except by kill -9. Which would make it an obvious bug: Not being supposed to interrupt a process that might be running for 24 hours on end is a pain. That said, you can abort port in relative safety while it's fetching files; and you can interrupt it while it's unpacking or configuring things provided that you "port clean" immediately afterwards. Killing it while it's building is more dangerous because you never know if the signal arrives in time. ;) So a slightly more useful signal handler might flag "abort requested", with port stopping once it's safe to do so. Regards, Marc