Re: [Qemu-devel] [PATCH] Permit zero-sized qemu_malloc() & friends

[Avi Kivity]

On 12/07/2009 07:40 PM, Anthony Liguori wrote:
> Avi Kivity wrote:
> > On 12/07/2009 07:17 PM, Anthony Liguori wrote:
> > > Avi Kivity wrote:
> > > > I don't understand.  People will develop patches for 0.12 for a 
> > > > while as bugs are reported and fixed.
> > >
> > > What's the exact problem here?
> >
> > Bug reported against qemu.  Developer develops patch, while testing 
> > qemu crashes on unrelated assert(size == 0).
>
> 1) Developer develops a patch against 0.12, it works, and they submit 
> it to upstream.
> 2) Upstream crashes because of assert(size==0).
> 3) Developer is publicly shamed for developing against a release 
> instead of a git tree.
>
> The problem is (1), not (2).   Not to mention that we won't allow 
> qemu_malloc() uses in upstream anymore which should make (2) impossible.

My problem is with stable-0.12.  Consider upstream fixed.

1) Bug reported against qemu-0.12.0.
2) Developer writes patch against master, submits, all is well except 
for the CODING_STYLE argument it triggers.
3) Developer writes patch against stable-0.12, can't test because 
testing crashes in some place where production doesn't crash.

The patch doesn't have to do anything with memory allocation.  By 
introducing gratuitous differences between developer and production 
mode, you're reducing quality.

--
Do not meddle in the internals of kernels, for they are subtle and quick to 
panic.