Re: oops? read/write vs type of length parameter

[Eli Zaretskii <eliz@gnu.org>]

> Date: Wed, 13 Apr 2011 09:06:14 -0700
> From: Paul Eggert <eggert@cs.ucla.edu>
> CC: jim@meyering.net, emacs-devel@gnu.org
> 
> On 04/13/2011 02:46 AM, Eli Zaretskii wrote:
> > I was asking whether testing errno for EWOULDBLOCK
> > and EAGAIN, and the code that deals with when that happens, are good
> > enough for all the possible reasons that emacs_write and
> > emacs_gnutls_write could return a partial count of bytes.
> 
> The revised send_process code tests for EWOULDBLOCK and EAGAIN under
> the same conditions that it does now.

That's true, but it's not my point.  The original code continued
looping and trying to write the rest of the stuff if emacs_write
returned a positive value that is smaller than what is was asked to
write.  The new code bails out of the loop in that case, assuming that
it's due to an error.

> > sysdep.c is not about system things, it's about Emacs code that
> > requires platform-dependent techniques.  Most of that indeed deals
> > with system types, but you will find quite a few Emacs specific
> > internals
> 
> Yes, and I had already mentioned that the abstraction boundaries were
> sometimes being violated there.  But let's not make matters worse.
> sysdep.c is supposed to be for interfaces to system-dependent kernel
> and library code, not for access to Emacs Lisp internals.

I was trying to point out that you are mistaken in your interpretation
of what sysdep.c is.  But even I accept your POV, access to OS-level
features does not mean we cannot use Emacs-specific data types there
or forget about Emacs design.  sysdep.c is not a library, it's part of
an application.

> >> For example, it would be a fairly small change to make Emacs buildable on
> >> a machine with 32-bit pointers and 64-bit EMACS_INT.
> > 
> > Somehow, I doubt it is a small change.  But if it is, by all means
> > let's do it now!  What are we waiting for?
> 
> I suspect you are joking, but if not, I'd be happy to implement it.

I'm not joking.  How could I?

> It shouldn't take that long, and it'd be nice to remove that silly 512
> MiB limit.  Would you object to such a change?

Of course not!  Why should I object to a great feature like that?

> in the meantime, we have a proposed change that
> fixes bugs and doesn't introduce bugs.  Since there are no downsides
> to this fix now, and no improvements to it have been suggested, we can
> install it now, and worry about future improvements later.

I'll let Stefan and Chong decide on that.  I still think that using
ssize_t as the last argument to emacs_write is better.

> > Either you care about future changes or you don't.
> 
> Both.  I care about plausible future changes; and I also realize that
> we don't have time to program for all possibilities and need to fix
> what's before us, which the proposed patch does.

IOW, you care for what supports your arguments and don't care about
the rest.