I wrote: > > Do you still expect GNULIB_TOOL_IMPL=sh+py to pass by the way? > > Sure. We need to give users a year, at least, for migrating from > the shell to the Python implementation. And when they migrate, > it should be a no-op for them.
Actually, no, it is not a requirement any more that GNULIB_TOOL_IMPL=sh+py works in all cases. This strict "same output" check was useful before we switched to the Python implementation by default, and in the few weeks afterwards. It's OK now to make small changes to the output of gnulib-tool.py, and change the unit tests accordingly.(*) This will make the unit tests fail with GNULIB_TOOL_IMPL=sh, but it is no longer important since the shell implementation is only maintained for a diminishing set of users. (*) Of course the changes to the expected outcome of the unit tests must be carefully reviewed! Bruno
