I've posted a bugzilla entry http://d.puremagic.com/issues/show_bug.cgi?id=10177 and finally managed to reduce to a small test case.

So will the proposed method improve error reporting?

Is there any case where it might hurt?

On Mon, May 20, 2013 at 3:24 PM, Timothee Cour <thelastmammoth@gmail.com> wrote:

In a number of cases I get very large compile error messages after running rdmd main_module.d

Upon inspection, running rdmd -v fun.d reveals that the irrelevant error messages are in different semantic3 passes:

semantic3 module_with_relevant_error
relevant error (eg: Error: no property 'x' for type 'y')
semantic3 module_with_irrelevant_error
irrelevant error (only coming from the fact that there was an error in a prior semantic pass)

...
and the list goes on for large number of lines.

Why not just start semantic3 passes at 1st error occurence ?
or at least only show those by default?

in all cases i've encountered all that was needed for me to see was the 1st faulty semantic3 pass, all other was irrelevant artifacts stemming from that.