Re: `message' not outputting the newline "atomically"

[Paul Eggert]

Eli Zaretskii wrote:
> We already found an agreed solution for
> 'message' and 'vmessage', so why step back and re-introduce them into
> the problem again?

Because we now have a simpler solution.

> As for other messages, I think the vast majority are diagnostics, and
> thus using stdout for them is inappropriate, as it will get in the way
> of using Emacs in scripts.

No, Emacs uses stdout reasonably extensively for this sort of thing already; see 
'write_stdout' and I can cite several other examples. In practice this has not 
been a problem because nobody cares whether these kinds of Emacs messages go to 
stdout or stderr. One can also see this in the decades-old commentary for the 
implementation of 'message', which says both 'stdout' and 'stderr' (obviously 
some of this commentary is incorrect, and the proposed patch fixes it).

All I'm suggesting is that we move a few messages from Emacs's stderr category 
to its (already-existing) stdout category, to fix the interleaving problem for 
messages that are not fatal diagnostics or for debugging. These messages are the 
ones most likely to cause interleaving problems in practice. The vast majority 
of messages would be unaffected by the proposed patch.

> what would happen if a script says
> 2>foo? will the file 'foo' remain empty if we use any stream but
> stderr inside message_to_stderr?

No, it'll work fine. The OS doesn't care about stdio's buffers; all it cares is 
whether file descriptor 2 is used, which it is.