Re: [Valgrind-developers] Stack traces with zero addresses in them...

[Nicholas N. <n.n...@gm...>]

On Thu, Feb 5, 2009 at 8:00 PM, Julian Seward <js...@ac...> wrote:
>
> On further consideration of this, a couple of comments:
>
> 1. Perhaps creating a new struct StackTrace is too much trouble.

Possibly... if it's variable-sized then you can't stack-allocate it
(can you?) and that would inconvenience many existing uses.

> 2. At the moment the machinery operates by passing around
>   (ips, n_ips) pairs.  Won't the proposed change cause
>   confusion about the meaning of n_ips?  Because then there
>   are two lengths to record: the size of the ips array, and
>   the number of actually used entries in it.
>
>   So maybe it's then relevant to consider a representation
>   (ips, ips_size, ips_used), where 1 <= ips_size <= VG_DEEPEST_STACKTRACE
>   and 1 <= ips_used <= ips_size ?

Yep.  I'll likely do something like this when I get to looking at it.

N