[DTrace-devel] [PATCH] Unify NULL pointer checking in dt_cg_check_notnull()

[Kris Van Hees]

On Fri, Jun 18, 2021 at 03:21:16PM -0400, Eugene Loh wrote:
> On 6/18/21 2:47 PM, Kris Van Hees wrote:
> 
> > On Fri, Jun 18, 2021 at 02:33:59PM -0400, Kris Van Hees wrote:
> >> On Fri, Jun 18, 2021 at 02:26:32PM -0400, Eugene Loh wrote:
> >>> Reviewed-by: Eugene Loh <eugene.loh at oracle.com>
> >>>
> >>> What will be the ordering of patches?  If this goes in first, I can
> >>> tweak dt_cg_act_[u]stack() to use it.  If those patches go in first,
> >>> this patch should include those two functions.
> >> This patch is planned to go before your stack related patches.  So, yes, you
> >> make make use of it.
> > Actually, what do you mean...  Where do you use NULL pointer checking in the
> > [u]stack implementation?  I do not see any...
> Good point.  Sorry.  It's not a BADADDR but a BADSTACK.
> 
> My main point is that by making the function broader, it could handle 
> all of these cases.  Indeed, one could eliminate the notnull() function 
> altogether.  Just beef dt_cg_probe_error() up to take a condition (NEQ, 
> SGE, etc.) and a register.  Thereby, one could serve the DIVZERO, 
> BADADDR, and BADSTACK cases, eliminating the boilerplate that currently 
> surrounds every dt_cg_probe_error() call.  To me, it seems cleaner than 
> using a narrowly defined notnull() function.

I'll give that some thought.  Definitely not happening right now.  I am also
not sure whether it will truly be an improvement or make for less clear code.

	Kris