[Castor215]

Closes #2557

As expressed in the issue, this needs feedback from other Linux users. Did pre-v148 ares not handle symlinks on your distro? Did the changes in v148 fix any issues? I don't want to cause any regressions so I'm opening this as a draft pending feedback.

[remutro]

I plan on reviewing behavior again. I think there is likely more changes that we might want to make even if this works, but I'll know more when I have some time to spend on it.

[bsdcode]

I can confirm v148 on FreeBSD is also affected. Didn't test previous versions. I couldn't test the proposed fix yet, but an additional || defined (PLATFORM_BSD) would probably be necessary.

[Castor215]

I plan on reviewing behavior again. I think there is likely more changes that we might want to make even if this works, but I'll know more when I have some time to spend on it.

Fair enough. If it helps, this is the problematic line in nall/nall/directory.cpp

158        if(readlink(result.data(), buffer, PATH_MAX) < PATH_MAX) {

readlink is returning relative links "as is" i.e. keeping them relative. The only way I can see around this is detecting if the first character is not /, finding the path of the link itself in that case and constructing the path contained in the link manually.

When execution returns to Emulator::load, you would need to detect if location is a symlink (absolute or otherwise) or not so the last used directory gets set to the symlink's location.

[bsdcode]

Fix is working on FreeBSD.