Ken Williams

@hofnerb about the previous comment - what happens if you move the ` - travis-tool.sh install_deps` line in `appveyor.yml` to be the last step in `build_script`, instead of the first...

So does this mean no fixes can happen on `mboost` for the foreseeable future?

Okay, thanks for the pointer. I'll check out `pak` - on first usage I'm getting some errors that I'll report in that project.

Hi @jimhester , this is an old thread, but I've been watching `pak` and it looks like it's kind of stalled out. The last release of it to CRAN was...

Do we know what actual change in `perl` or `File::Spec` caused the difference in behavior?

> Interestingly `#if defined(HAS_LSTAT) && defined(HAS_READLINK) && defined(HAS_SYMLINK)` which was `FALSE` in perl-5.32.1 and earlier, is `TRUE` in 5.34.0 and later. That might be the most interesting fact, actually. Perhaps...

> TBH I think this needs to get fixed on the Path::Class side. Not IMO. I think the `Path::Class` test exposed a regression in either `perl` or `Cwd` that needs...

> I think you voided the warranty the moment you did BEGIN { $^O = 'Unix' }. The obvious solution is to change the tests to use File::Spec instead of...

An observation - it appears that `File::Spec` does intend, at least somewhat, for users to be able to work with logical paths as represented on other OSes (see https://metacpan.org/pod/File::Spec#DESCRIPTION): >...

Very interesting - so apparently `Cwd::cwd()` doesn't care about `$^O` being wrong on Windows (which seems kind of weird to me!), but `Cwd::realpath()` does. If I scanned through the code...