Possibly back out of #125 and #115

[oalders]

I think we need to back out of https://github.com/libwww-perl/HTTP-Message/pull/125 and https://github.com/libwww-perl/HTTP-Message/pull/115
5:01 PM 115 introduced problems which 125 tries to fix, but can't completely remedy
5:01 PM So we get more failures than in earlier releases
5:02 PM anyone opposed to reverting those PRs and *then* releasing?
5:03 PM As I recall, to debug this we'd need someone on a Windows machine and using the charset that triggers the errors

References PRs: #115 and #125

[oalders]

I don't think we can release a new HTTP::Message without dealing with this.

[oalders]

@genio assigned ticket to you as we had discussed Windows testing on IRC.

[genio]

IIRC, we are trying to create a Windows-friendly filename right now and we're coming from a UTF-8 world -which Windows isn't when it comes to the file system.

What I think we tossed around was something akin to:

if ($^O eq 'MSWin32') { # possibly should do this in Cygwin as well
    try {
        require Win32;
        my $fname = Win32::GetANSIPathName($unicode_file_path);
        # do something useful with our windows-compliant filename here
    }
    catch {
        die("We had a problem getting the appropriate Windows-compatible filename: $_");
    };
}

I'd need to review that area a bit further to ensure that hand-wavy bit up there is correct.

[genio]

@vanHoesel @skaji @karenetheridge Does that make sense? ^^

[karenetheridge]

Sure, why not send that out as a -TRIAL release and see what the smokers say?

[oalders]

Changes have now been reverted.