2022年12月7日(水) 8:40 L A Walsh <b...@tlinx.org>: > [...] Similarly w/bash -- > I haven't tested it, but I'd expect bash compiled for windows > (using mingw toolchain) to use CRLF, but LF for WSL or Cygwin.
I think there is actually no Bash compiled for Windows (i.e., the pure Windows API on the Windows subsystem). The Bash that comes with the MinGW toolchain is linked with msys-2.0.dll (in the case of MSYS2), which means that the POSIX layer Bash relies on is provided by MSYS which is a minimized fork of Cygwin. The MSYS Bash treats LF as the newline but not CRLF. > Are you using both tools for the same OS and subsys and having > them conflict? I think so. I think this means that the reported configuration is wrong or, at least, very unusual. I don't think we should add in Bash an option that is only meaningful in a specific non-unix-like operating system for a heterogeneous amalgam of programs from different subsystems. That option is practically useless in all of the major Unix-like systems. If something would be modified at the side of Bash, maybe there is a chance that Bash of the Cygwin/MSYS packages could be patched like `shopt -s completion_strip_exe'. But even with that case, a question is why filtering by `tr' is not an option. The answer seemed to be to make the program work unmodified, but I don't think we should expect that the combination of programs from different subsystems will work unmodified in general. -- Koichi
