Modernizing core code (was Re: [Nmh-workers] tempfile creation

[heymanj]

On 27 April 2005 at 13:41, Neil W Rickert <address@hidden> wrote:

> The real problem here is that recent nmh "development" has been in
> the form of creeping featurism.  But the code base is rotten to the
> core, and badly needs revamping and modernizing.  Some documenting
> would also be in order.

As a long time user of MH (I still have Jerry Peek's 1st edition of the
classic MH & Xmh), I've ported it to several exotic Unix platforms when 
I've needed to.  Never could figure out who to get the code changes back
to - and of course now they're gone :-(

Recently attemped to port 1.1RC2 to QNX Neutrino 6.2.  Why you ask?
For lack of anything better to do at the time :-)  Unfortunately, nmh
doesn't port easily to non-Unix systems, even those that provide a POSIX
compliant API set.  My biggest hurdle, which lead to my abandonment of
the porting effort, was the fact that the code uses intimate knowledge
of stdio.h.  More specifically, m_getfld.c, is full of porting issues:

/* 
   This routine "knows" that struct FILEs have a _ptr
   and a _cnt to describe the current state of the buffer and
   it knows that _filbuf ignores the _ptr & _cnt and simply fills
   the buffer.  If stdio on your system doesn't work this way, you
   may have to make small changes in this routine.
*/

And trying to map QNX's FILE definition to that used by nmh became
too much of a nightmare.

As a question, is it worth attempting to abstract this out, so this
particular pitfall doesn't exist in the future?  It would make porting
nmh to non-Unix systems significantly easier in the future.   Several 
times I have thought about it - and even posted to what I thought was 
the right place to get on this list - only to never receive a response...

> 
>  -NWR

jerry
       //  Jerry Heyman      | "Software is the difference between
      //   Amiga Forever :-) |  hardware and reality"
  \\ //    address@hidden   | 
   \X/