Bug 3457 – rdmd fails silently in a particular setup where the compiler is not the expected

I have a set up with both DMD1 and DMD2. I call the binary for DMD1 dmd and the binary for DMD2 dmd2. When running rdmd2 (rdmd for DMD2 ;) it returns 1 to the OS and nothing happens. The problem seems to be rdmd2 is trying to use dmd as the compiler and it fails to include some D2 specific file: $ rdmd2 --chatty ./buildit.d chdir '.' && dmd -v -o- './buildit.d' >./buildit.d.deps $ echo $? 1 $ dmd -v -o- ./buildit.d parse buildit semantic buildit import object (/home/luca/tesis/dmd/linux/bin/../../include/object.di) import std.path (std/path.d) ./buildit.d(3): module path cannot read file 'std/path.d' I know this is a weird setup, and that it's OK that rdmd fails, but it would be helpful to have some indication about what happened, like: Failed to execute: dmd -v -o- './buildit.d' >./buildit.d.deps The weird thing is when using dmd2, if the compilation fails, rdmd shows the dmd error messages perfectly, so I don't know why it's eating the error messages using DMD1 as the compiler.

This is a bug in dmd; it doesn't print that particular error message to stderr, it prints it to stdout. I submitted a bug report to dmd and wrote a workaround in rdmd.