< vasild>
How to observe debug.log after a run of test_bitcoin? It gets deleted at the end of the execution. To see what is inside I did run it inside gdb, set a breakpoint in unlink and when it stops at the breakpoint: cat /tmp/test_common_Bitcoin\ Core/*/debug.log
< vasild>
"test_bitcoin -- -debuglogfile=/dev/stdout" did not print the log on the screen, nor did -printtoconsole=1
< bitcoin-git>
[bitcoin] ccdle12 opened pull request #19350: test: Refactor tests using restart_node (master...test-refactor-restart-node) https://github.com/bitcoin/bitcoin/pull/19350
< jonatack>
vasild: for temporary printing with unit tests, so far i've managed with std::cout or std::cerr or BOOST_TEST_MESSAGE() combined with running the tests with --log_level=all
< vasild>
jonatack: right, I also use printf() or std::cout sometimes, but some stuff is printed to debug.log by non-test code and I was wondering if that can be observed
< jonatack>
yes, sorry, just realised i didn't answer your question
< vasild>
I vaguely remember some --dont-remove-logs-at-the-end option, but maybe it was in some different software :)
< fanquake>
vaslid: can you use --log_sink=stdout?
< vasild>
trying...
< fanquake>
You can also control verbosity with --log_level. Anything between 'all' and 'nothing'
< fanquake>
Quite a lot of info in --help if you haven't looked through it already
< jonatack>
yes, the help is good, as well as the details of --help=<option>
< vasild>
(I had to recompile, sorry for the delay)
< vasild>
fanquake: --log_sink=stdout is the default, it does not help
< fanquake>
I thought it might be
< vasild>
I think the deal is that --log_level and --log_sink are boost-test-framework options and control the outputs from it. However bitcoin code unrelated to that prints some stuff to datadir/debug.log
< vasild>
This helps:
< vasild>
- "-printtoconsole=0",
< vasild>
+ "-printtoconsole=1",
< vasild>
in src/test/util/setup_common.cpp
< vasild>
but it involves modifying the code
< bitcoin-git>
[bitcoin] hebasto closed pull request #18307: build: Require pkg-config for all of the hosts (master...20200309-pkgconfig) https://github.com/bitcoin/bitcoin/pull/18307
< bitcoin-git>
[bitcoin] hebasto reopened pull request #18307: build: Require pkg-config for all of the hosts (master...20200309-pkgconfig) https://github.com/bitcoin/bitcoin/pull/18307
< sipa>
(which means it's ARM, because ARM doesn't produce its own hardware)
< provoostenator>
Hopefully their automagic conversion doesn't completely ruin our performance, but apparantly it might be 2 years before anything ships, so
< provoostenator>
They make it possible to compile a single binary for both Intel and Apple CPU.
< provoostenator>
So we can do nothing, or use that new binary type.
< provoostenator>
Let's see how their security holds up...