Jaromír Doleček <dolecek@ibis.cz> writes:
> Is the amount of "extra maintanence work" really so bad to warrant
> addition of the *DEFINE() hackery ? Doesn't that boil down just to
> "they do it differently then what we are used to" ?

having to provide diffs (rather than a tarball of files, or even just
.o's) is significant extra work, yes.  and, maintaining those diffs is
annoying, time consuming, and means that time T down the road your
code is that much less likely to drop in.

It's certainly bad enough to warrant a better dynamic type
registration scheme.


> I kind of like
> when globally used information is kept centralized.

That's the thing, though.  w.r.t. malloc types, the information is
used globally only in a couple of ways:

	* tools which read out the data, and

	* tools which do global accesses of the form 'array[index]',
	  where index is the malloc type.

with a few exceptions, use of each malloc type is (or should be)
limited to a relatively small set of code.

similarly, how many things globally use the VFS list?  why, that'd be
sys_mount().  8-)


> BTW, is it really needed for malloc ? Shouldn't the code use
> pools for "permanent" stuff and M_TEMP for temporary ? ;-)

pools are for fixed-sized objects.


cgd