hmm.. Re: talloc issues

[Sam Liddicott]

* tridge wrote, On 28/07/09 12:00:
> The point of my changes was that the review would happen automatically
> and be absolutely accurate as we are catching all the cases where
> there are problems using the logic in the code. The only places where
> semantics change are all caught by the warning I added. So if no
> warning is produced then we know the code is doing something that is
> OK.
> 
> For Sam's proposed changes there are no such guarantees. All code
> needs to be manually reviewed, even the ones that seem naive on first
> glance. When I added the warning to my changes it took Andrew and me
> quite a while to even find the cause of the warning in some cases, as
> they were spread across multiple files. Some of them looked "obviously
> correct" but they weren't.

I accept your claim that manual examination won't find all cases.

I am surprised that talloc_free is used in so many places, and think
this is what needs removing;

However I'm forced to accept your judgement that with current code, some
cases are too hard for the human mind to understand.

I'm still left with the problem that the resulting talloc is not useful
to me.

Sam