use MALLOC_PERTURB_ ... or lose

[Jeff Spaleta]

On Wed, May 5, 2010 at 9:28 AM, Eric Sandeen <sandeen at redhat.com> wrote:
> Agreed, I'm tired of (insert random benchmarking site) saying "OH NOES!
> Fedora got SLOWER AGAIN!" when it's really a lot of debug going on.
>
> Stating something like this clearly on login & install would be nice,
> not just for this MALLOC_PERTURB_ change but in general.

Doesn't this whole discussion about debugging versus performance also
apply to F13 pre-release testing as well.. and not just rawhide?  At
what point do we turn off additional debugging in the kernel and
whatnot ahead of going gold in the new branched pre-release workflow?
Is the kernel debugging still on in F13 systems right now?  If there
really is a debugging win here should we sync the MALLOC_PERTURB
enablament policy to whatever is happening with kernel debugging for
branched pre-releases as well as rawhide?

This way pre-release systems installed at the time of the beta have
debugging enabled, and users are informed of the debugging effort (and
thanked for their participation!), but at some point via an updated
package it gets turned off (barring any local overrides) a some
definite point in the pre-release run up.


-jef