< luke-jr> achow101: shouldn't you be able to invalidateblock at least? (before upgrading)
< achow101> luke-jr: I don't think so. from what I have seen (of other people's nodes experiencing this problem), DisconnectBlock chokes on disconnecting the invalid block.
< achow101> also, the most work chain on testnet is a valid one, so old nodes should automatically reorg to that one, if they weren't getting stuck on DisconnectBlock
< gmaxwell> luke-jr: no because it cannot disconnect it.
< gmaxwell> It tries to restore the earlier spend and find that its already there.
< gmaxwell> and doesn't know how to undo the duplicate.
< gmaxwell> the startup sanity check also detects this case.
< luke-jr> ah
< luke-jr> which also means a reorg will fail I guess?
< gmaxwell> right.
< gmaxwell> And it does.
< gmaxwell> It just get stuck at the bad block.
< gmaxwell> and if you restart, the startup check will tell you to reindex.
< echeveria> sort of funnily, it means that on reorganising to the non-invalid chain, your node is disabled. seems like an accidental win-win to me.
< gmaxwell> It does keep you from contributing to more issues, and makes sure you'll upgrade.
< fanquake> wumpus: thanks
< jonasschnelli> Seems one of the travis jobs don't make it (repeatedly) under 50mins: https://travis-ci.org/bitcoin/bitcoin/jobs/435205266
< Varunram> just curious, why doesn't core have multiple tests on various jenkins/ other CI servers instead of relying on travis (which seems to break often?)
< ken2812221> #13515 is a solution to save cache before timeout
< gribble> https://github.com/bitcoin/bitcoin/issues/13515 | travis: Avoid timeout without saving caches, also enable qt builds for all jobs if available by ken2812221 · Pull Request #13515 · bitcoin/bitcoin · GitHub