bug#51327: 28.0.60; emacsclient warns about XDG_RUNTIME_DIR when starting daemon on demand

[Jim Porter]

On 11/5/2021 3:38 AM, Ulrich Mueller wrote:
> If I understand this report correctly, the problem is just the spurious
> warning about XDG_RUNTIME_DIR?
>
> Instead of changing the functionality (which breaks other use cases, see
> my message to emacs-devel), wouldn't it make more sense to just suppress
> the warning if the variable is set? As in attached patch?

It's not just a spurious warning; the warning is telling the user about 
a real problem, though the wording is a bit confusing for this 
particular case. If a user calls `emacsclient --alternate-editor=""' 
with XDG_RUNTIME_DIR set and no Emacs server running, emacsclient will 
check in both XDG_RUNTIME_DIR and TMPDIR to find the server socket 
before giving up and starting the daemon.

Since XDG_RUNTIME_DIR exists (at least in part) to prevent symlink 
attacks, Emacs should try to avoid checking TMPDIR in order to avoid 
this vulnerability. Emacs 27 is secure in this regard, since it *never* 
checks TMPDIR if XDG_RUNTIME_DIR is set. However, that behavior caused 
the problems described in bug#33847. The patch I posted is a compromise 
that restores the secure behavior for users who set the alternate editor 
and want to start the Emacs daemon on demand (it's not perfect though; 
see my reply in emacs-devel).