Marcus Müller

Can't reproduce right now, no reliable windows at hand, will ask for help

@willcode no API change, only ABI-irrelevant annotations, and additions: designed for backportability.

oha. so, situation: - because the parent directory doesn't inherently exist, need to create ~/.config/gnuradio recursively - Windows is fun: https://en.cppreference.com/w/cpp/filesystem/path/generic_string ; need to `path::string().c_str()` to get the path in...

TODO: check whether CI runs all fail during testing or whether something strange is going on; some fail surprisingly fast.

I don't have any Mac OS machine to reproduce this – help from someone with one utterly needed. Link to repro flowgraph: https://gist.github.com/marcusmueller/eb13db71a2b7b3b95c020f703722fc98 

I think this was closed in #6697 ; running valgrind in the background, but `--leak-check=full` is not very fast.

Yeah, aside from the inherent Python leaks, valgrind now looks clean to me!

ah, I added that tag so I don't forget to look at it when valgrind was finished, and then it didn't take as long as I thought. Here's how I...

Interactively, however, yes: ``` ==74295== 88 bytes in 1 blocks are definitely lost in loss record 4,154 of 5,841 ==74295== at 0x4842F95: operator new(unsigned long) (vg_replace_malloc.c:483) ==74295== by 0x6B300E4: gr::network::tcp_sink_impl::connect(bool)...

To make this easier on future readers, this is the offending code: ```c++ asio::ip::tcp::socket* tmpSocket = new asio::ip::tcp::socket(d_io_context); d_acceptor->async_accept(*tmpSocket, [this, tmpSocket](const asio::error_code& error) { accept_handler(tmpSocket, error); }); ```