Re: attribute warn_unused_result

[Paul Eggert]

On 02/06/2011 11:27 AM, Eli Zaretskii wrote:
> I actually question the resulting behavior on Posix platforms: AFAIU,
> Emacs will (almost silently) do nothing.  Is that a good UI?  Should
> we perhaps signal an error instead?

Emacs can't merely signal an error there, since it's in the child.

My recent change causes Emacs to report chdir failure the same way
it reports execlp failure.  If the former behavior isn't right,
the latter isn't either; and I expect that the latter behavior has
been there for decades.

For modern mainstream platforms the issue doesn't come up, as
this code is exercised only on weird POSIXish platforms that
can't suspend Emacs.  I last used a platform like that in
the 1980s, and I expect they're very low priority now.

For Windows things may well be different, but that
difference can be put into #ifdef DOS_NT code.