With every log-in: realpath() cannot resolve path ""

[rdmark]

Since the change in #212 I'm seeing this once every time a user connects to an afp volume:

afpd[15035]: realpath() cannot resolve path ""

So there is code somewhere that tries to resolve an null string path. Doesn't seem fatal, but arguably a bit unclean. While figuring this out, perhaps lowering the log level to warning is apt.

[rdmark]

To tweak the log level: #279

[dgsga]

The relevant code is present in the realpath_safe() function, starting at line 438 of libatalk/util/unix.c. Realpath_safe() is called in etc/afpd/fce_api.c and libatalk/util/netatalk_conf.c...

[rdmark]

Thanks for tracking down the usage of the code. I narrowed it down to etc/afpd/fce_api.c where fce_register() is called with an empty string for path in etc/afpd/auth.c.

netatalk/etc/afpd/auth.c

Line 289 in aca367a

fce_register(obj, FCE_LOGIN, "", NULL);

That empty string is then passed to realpath_safe()

In fact, there are two places in auth.c where fce_register() is called with an empty string: once for login once for logout.

Elsewhere in the codebase, fce_register() is always called with a path string, so the issue seems to be isolated to this spot in the code.

[rdmark]

Found a simple fix: move the path lookups closer to where they're actually used in that function. This allow fce_register() to validate if listeners are configured, which they aren't at this stage, and return with AFP_OK before the path lookup.

#292