< bitcoin-git> [bitcoin] BitcoinErrorLog opened pull request #19416: Update pruning tooltip (master...patch-1) https://github.com/bitcoin/bitcoin/pull/19416
< bitcoin-git> [bitcoin] Saibato opened pull request #19419: wallet: Listwallets do not iterate trough our blocksdata. (master...wallet_351) https://github.com/bitcoin/bitcoin/pull/19419
< bitcoin-git> [bitcoin] fanquake closed pull request #19416: Update pruning tooltip (master...patch-1) https://github.com/bitcoin/bitcoin/pull/19416
< luke-jr> trying to figure out how to use Guix, but the instructions I find expect me to mess with my system/root; is there a howto do it within a single isolated directory, like gitian?
< luke-jr> (or is moving to Guix a bad idea after all?)
< luke-jr> according to #guix, Guix itself has bootstrapping problems.. wtf
< luke-jr> smh, dongcarl correct me if they're wrong
< promag> is it possible to run a functional test in lldb/gdb?
< luke-jr> promag: I use this
< luke-jr> while ! gdb -p `ps -eo pid,cmd | grep bitcoind.*testnode2 | grep -v grep | cut -d" " -f1` ; do true; done
< luke-jr> usually quick enough
< luke-jr> take care to adjust the node number
< promag> not sure that will work in my case
< promag> while test/functional/feature_shutdown.py; do :; done
< promag> eventually it crashes
< luke-jr> hmm
< luke-jr> does it dump a core file? ;)
< gwillen> promag: I have edited the python to put a long "sleep" at the top of the offending test so I have a chance to attach the debugger
< gwillen> I guess that doesn't work if you have to run the test in a loop though
< gwillen> the test is simple enough you might be able to reimplement it as a shell script and invoke the debugger that way
< promag> gwillen: yeah, I tried that but I feel I should be able to debug with the functional test - what if the test was complex?
< promag> luke-jr: right, I don't see the dump
< gwillen> unfortunately coredumps are now a function of systemd (I wish I was kidding about that)
< gwillen> check the value of 'ulimit -c' to see if you have a process limit on core dump size (0 means no core dumps)
< gwillen> otherwise there are some systemd configs involved, and allegedly by default the dumps should end up in /var/lib/systemd/coredump, but apparently you have to use `coredumpctl` to read them.
< promag> im on macos
< promag> ive done ulimit -c unlimited
< luke-jr> gwillen: lolwtfbbq?
< luke-jr> gwillen: yet another reason not to use systemd
< promag> so can we bump min libevent to 2.1.x?
< bitcoin-git> [bitcoin] promag opened pull request #19420: http: Track active requests and wait for last to finish (master...2020-06-track-requests) https://github.com/bitcoin/bitcoin/pull/19420
< jnewbery> promag: not sure if this helps in your scenario, but have you read https://github.com/bitcoin/bitcoin/blob/master/test/README.md#attaching-a-debugger ?
< jnewbery> also take a look at --pdbonfailure to attach a python debugger (and essentially pause the test) when an assert fails
< instagibbs> jnewbery, i dont find pdbonfailure too helpful because you lose context of the failure, at least when i tried it
< instagibbs> ideally it would just literally stick a pdb.set_trace() in the spot you hit an assert failure and let you "step" over the failure, or something
< jnewbery> instagibbs: right, you're not inside the main run_test() function any more, but you can still interact with the nodes eg run self.nodes[0].getblockchaininfo() or whatever, or attach a gdb to the bitcoind process
< instagibbs> right depends on the issue you're debugging, can be helpful
< jnewbery> I guess we could change the assert_***() functions to attach a pdb _before_ asserting if you still want the run_test() context
< troygiorshev> PSA - As of recently, clang-format now requires version 9 or greater. Recent (19.04+) ubuntu releases are already there. For most others upgrading should be easy, just search for clang-format-9. If you use debian, go here https://apt.llvm.org/
< troygiorshev> This is due to a recent change to our .clang-format file
< jonatack_> troygiorshev: thanks, do you know whch PR or why was this changed? I'm on debian clang 6 and you are right, for the first time clang-format is broken now
< troygiorshev> jonatack_: it's the addition of AllowAllArgumentsOnNextLine : true in 19095.
< troygiorshev> #19095
< gribble> https://github.com/bitcoin/bitcoin/issues/19095 | [tools] Update clang-format config for multi-line function declarations and calls by jnewbery · Pull Request #19095 · bitcoin/bitcoin · GitHub
< troygiorshev> also stretch is EOL in 6 days, go upgrade to buster :D
< jonatack> troygiorshev: yeah, i use a security-by-default variant of debian with coreboot, that is conservative with upgrading and it's still at clang 6 for now. last time i tried to install llvm from source i wasted a lot of time without managing to.
< jonatack> oh, well
< troygiorshev> jonatack: https://apt.llvm.org/ should still work nicely for you (give or take security considerations)
< jonatack> it does not
< jonatack> i'll live without clang-format for now
< bitcoin-git> [bitcoin] MarcoFalke opened pull request #19422: ci: Add tsan suppression for race in wallet (master...2007-ciTsanSupW) https://github.com/bitcoin/bitcoin/pull/19422