Re: Dumper issue, revisited; invalid realloc/free

[Rich Felker]

On Thu, Feb 05, 2015 at 05:14:54PM +0100, Wolfgang Jenkner wrote:
> On Wed, Feb 04 2015, Rich Felker wrote:
> 
> > I now have a working build with no ugly hackery.
> 
> Does it use any special properties of musl malloc or should the hybrid
> approach now work with any reasonable malloc implementation?

No special properties at all; that's what I meant by no ugly hackery.
In fact it seems like it _couldn't_ depend on anything about the
malloc implementation, because the system malloc implementation is not
used prior to dumping and the pre-dump objects are never passed to it
after dumping.

So, based on both the concept and the practical success with it
(everything worked basically first try with no difficulty on a new
target), I believe hybrid malloc should be the default for platforms
where there's not a reason to prefer something else.

Rich