rd_assert on rd_buf_alloc_segment0() call

[mortenjc]

I am a happy user of librdkafka for a data acquisition system for a scientific institution.

Due to a recent server re-configuration ulimits were changed from unlimited memory per process to 4GB. My application
seemed to need a bit more (5.4GB).

This caused the rd_assert() I presume - see the backtrace

and subsequently the whole application crashed.

This seems a bit harsh to me. Has anyone experienced this and is there some way to avoid it (except removing the limits of course)?

Would it perhaps be useful for the application to get notified that the produce call was unsuccessful instead so at least it has a chance of failing more gracefully? Or is there something fundamental I am missing?

Cheers
Morten