On Tuesday 24 April 2007 16:24, Philip Hazel wrote:
> On Tue, 24 Apr 2007, Magnus Holmgren wrote:
> > Perhaps the best (or simplest) solution is, after all, to add
> > body_linecount and body_zerocount to the API, and make it the
> > responsibility of the local_scan() function to update them?
>
> I suppose so, though I do not really want to encourage people to write
> to the fd. However, given that they can see these variables anyway, via
> the "back door" I suppose it isn't giving away much to put them into the
> official API, though it's a bit of work to update the code and the
> documentation.
Except in Debian, where we've eliminated the symbols that aren't included in
the API from the dynamic symbol table.
> > Isn't it true that all line endings are converted to '\n' on
> > reception?
>
> Yes, that's true.
So my patch should be correct, except that perhaps message_size and
message_body_size need not be updated as they will be updated before delivery
anyway?
I suppose that APIs shouldn't be changed unnecessarily. On the other hand,
I've often found the local_scan API too limited, letting local_scan() update
the variables is more efficient, and in the case of SA-Exim (the only
publicly available local_scan module I'm aware of), it means that the *de
facto* API stays the same.
If the API is to change, I suggest that it be looked over and other potential
extensions be made at the same time (for example, a way to build strings
efficiently, instead of with s = string_sprintf("%s foo bar", s), has been
requested. We should not say "Exim will update body_linecount as needed" and
then change that to "You have to update body_linecount yourself".
> > Whatever the decision, I'd very much appreciate one in the next few days,
> > otherwise I'll have to ask Marc Haber to back out the "reduce the dynamic
> > symbol table" patch from the "dynamic local_scan" patch.
>
> Well, there won't be another Exim release for some time, given that 4.67
> came out last week. But I guess you can patch it. :-)
Marc Haber does that all the time. Right, Marc? :-)
What do you suggest? Until this is definitely sorted out, should I upload an
sa-exim that doesn't update body_linecount, or should you upload an exim4
that doesn't reduce the dynamic symbol table?
--
Magnus Holmgren holmgren@???
(No Cc of list mail needed, thanks)
"Exim is better at being younger, whereas sendmail is better for
Scrabble (50 point bonus for clearing your rack)" -- Dave Evans
