[erlang-questions] Tracing and debugging
Richard A. O'Keefe
Fri Jan 16 04:05:40 CET 2015
On 14/01/2015, at 2:57 am, Richard Carlsson <carlsson.richard@REDACTED> wrote:
> What really should be done is that the current interpreting debugger should be destroyed with fire and a real bytecode-level debugger should be implemented (with low level functionality done in C), which would be way faster, only require updating when new Beam operations are added, and not require the abstract code to be present in the beam files.
There are certainly advantages to a byte code debugger,
but byte-code compilation destroys information like variable names
and source locations.
The thing I want above all from a debugging tool is that it should
minimise the amount of output I have to look at. I want to be
able to tell it my expectations and have it only show me things
that are surprising. Given the amount of filtering and suppression
of repetition that are appropriate to economise on *my* time, it
is not clear that working at the byte code level *would* be any
faster over all.
That is, after all, one of the Erlang Lessons: speed is a property
of whole systems, not just of compilers.
More information about the erlang-questions