Results 124 comments of Arthur O'Dwyer

> My only reservation with this PR is that this would result in all bytes being copied (regardless of `size()`)... > This has always been the tradeoff. I prefer triviality,...

> @Quuxplusone thanks for the description of the problem. Will try to reproduce it, maybe it's linked to stopping and restarting the broker. Could you try it without `rd_kafka_conf_set_rebalance_cb` to...

> @Quuxplusone we're also facing similar issues in our use-cases, but it's very sporadic. We're working on a stable reproducer to move further. Checking if you have a public reproducer...

> * that all these objects are destroyed before calling `rd_kafka_destroy()`, even in corner cases > > > For C, make sure the following objects are destroyed prior to calling...

Gentle ping! and question: what's the best way to deal with the current test failure, which indicates that you can't use `override` in C++98 mode? I'd argue that C++98 is...

Updated! I named the new macro `RD_OVERRIDE` for consistency with `RD_EXPORT`. clang-format seems mostly happy with it; I tried also adding `AttributeMacros: ['RD_OVERRIDE']` to the clang-format file, but that (A)...

Rebased — post-Thanksgiving ping! I think it would be nice to add `-Wsuggest-override` to the project's default CXXFLAGS as well (when Clang is being used), just to avoid regressions in...

I've rebased the patch. We still haven't resolved [what is the intended semantics of `fail(...)` in this example](https://github.com/confluentinc/librdkafka/pull/4114#discussion_r1051214093). Being a C++ chauvinist myself, I prefer to think of `fail(...)` as...