zsh and portability

zsh and portability

[Richard Coleman]

There been a lot of talk in the last couple of days about
various assumptions that zsh makes (NULL has all zero bits,
or char has 8 bits, etc..).

My suggestion would be to not worry too much about such things unless
they become a problem.  I want zsh to be portable as well, but not at
the expense of needlessly complicating the code for machines that zsh
will probably not run on anyway.

In any portable unix software, you must make assumptions that are not
guaranteed by any standard (ANSI C, POSIX, whatever).  I don't think
anything is wrong with this (withing reason).  If the code is kept
clean and well documented, and someone REALLY wants to run zsh on an
OS that goes against these assumptions, then they could can port it
themselves without too much trouble.

I think time could be better spent documenting the assumptions that
are made, rather than adding more complexity in the name of
portability.

Remember.  Keep things as simple as possible.  I will of course,
repeat this mantra on a regular basis.

rc

Re: zsh and portability

[Zoltan Hidvegi]

> There been a lot of talk in the last couple of days about
> various assumptions that zsh makes (NULL has all zero bits,
> or char has 8 bits, etc..).
> 
> My suggestion would be to not worry too much about such things unless
> they become a problem.  I want zsh to be portable as well, but not at
> the expense of needlessly complicating the code for machines that zsh
> will probably not run on anyway.
> 
> In any portable unix software, you must make assumptions that are not
> guaranteed by any standard (ANSI C, POSIX, whatever).  I don't think
> anything is wrong with this (withing reason).  If the code is kept
> clean and well documented, and someone REALLY wants to run zsh on an
> OS that goes against these assumptions, then they could can port it
> themselves without too much trouble.
> 
> I think time could be better spent documenting the assumptions that
> are made, rather than adding more complexity in the name of
> portability.
> 
> Remember.  Keep things as simple as possible.  I will of course,
> repeat this mantra on a regular basis.

I agree, but if a change is simple and trivial, and improves portability
than it would be usefull to add.

Also I think that in the near future there will be systems where
sizeof(long) > sizeof(void*).  The other problems will probably not come up
on modern systems but this one has to be handled somehow.  In other words
we need a portable method to initialise a union.

Zoltan