[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Status of PL48 Was: gcc-as 2.5.1



> Also, err.h is none exitant at all, so for some port that needed it I
> had to transform all the alertx() type into printf(),  It is not very
> cool, but well it works, anyone ok to port a err.h for mint? that
> would make some bsdi port much easier and faster.

Perhaps you should try to convert alertx() to S_alert() - which is a
regular kernel function - rather than to printf()?

> Also, my termcap.h seems to have a problem with all the tput()
> function. It compiles but with tons of warnings.

As I told you once, warnings can be successfully ignored if the programmer
knows what he's doing. A warning message is something different, than an
error message. The error means, that there's a serious mistake and the
source code can't be compiled and/or linked because program modules are
inconsistent. A warning message has a different purpose. Namely, it is
intended to warn the programmer, that the construction he has used MAY be
improper or not portable. Especially, the compiler becomes nervous and
produces warnings on some tricks or simplifications those have been used
by the programmer to make to object code shorter/faster on destination
machine. For example, we all know that on Motorola 68k a pointer is a 32
bit integer value. So we can assign such a value to a pointer, what in the
machine code looks like this:

	move.l d0,a0

However, if you try to compile a program which contains a part like this:

	char *cp;
	register long var;

	cp = var;

the last line will produce a code like this:

	move.l d0,a0

and a warning like this: "assignment makes pointer from integer without
a cast". If you replace the line with:

	cp = (char *)var;

you'll get rid of a warning, but the produced code will be the same, i.e.

	move.l d0,a0

In other words, warning messages may be meaningless, as you can
successfully write a program which will compile without any warning, but
it won't work either (causing bus errors or segfaults for example). At the
other hand, another program may produce hundreds of warnings, but may work
extremely fine (like MiNT kernel does).

> KGMD installation all gem based yes, but not as regular use.

It is not worth spending weeks on developing a GEM based installer, if
installation scripts do the same fine and they're much more easy to
change/fix when necessary. At least, most people install the system once,
so no need to spend time on visual stuff only. Its better to spend this
time on improving the system itself, and if the system itself will be
perfect, people will tolerate a non-GEM installer. Win95 has fine, graphic
installer which is sophisticated enough to work a hour without a crash and
looks fine. The other thing is that Win95 users may watch the installer
even once a week or more, so m$ had to make it looking fine, because the
program is used very often (I wish it played a movie while installing, it
might me much more attractive and less boring then).

Konrad M.Kokoszkiewicz
mail: draco@mi.com.pl
http://www.orient.uw.edu.pl/~conradus/
http://www.obta.uw.edu.pl/~draco/

** Ea natura multitudinis est,
** aut servit humiliter, aut superbe dominatur (Liv. XXIV,25)
*************************************************************
** U pospolstwa normalne jest, ze albo sluzy ono unizenie,
** albo bezczelnie sie panoszy.