Re: Gnus issues

[Simon Josefsson <jas@extundo.com>]

Dave Love <fx@gnu.org> writes:

> It seems it still uses nnheader-accept-process-output, which it
> shouldn't.  A reasonable solution to that, which I should probably
> have used elsewhere before, is to compile it away.  nnheader could
> just be required at compile time if nnheader-accept-process-output is
> made a defsubst (with a comment about why).

n-a-p-o uses a variable, nnheader-read-timeout, so if it is a
defsubst, wouldn't either the user lose the ability to customize the
variable, or that nnheader.el is loaded for the variable?

Btw, I should installed a fix to make pop3.el not depend on nnheader.
I copied the nnheader code, which may be bad, though.

OTOH, I'm not sure why we bother.  The entire n-a-p-o stuff seem to be
workaround for a problem on non-unix platforms.  Wouldn't it be better
if the people using those platforms fix accept-process-output to do
the right thing, instead of messing up the lisp code?

I would support using accept-process-output directly in pop3.el.