[Secure-testing-team] inject-embedded-code-copies

[Michael S. Gilbert]

On Wed, 26 Aug 2009 21:04:08 +0200, Moritz Muehlenhoff wrote:
> On Wed, Aug 26, 2009 at 02:55:03PM -0400, Michael S. Gilbert wrote:
> > On Wed, 26 Aug 2009 20:24:36 +0200, Moritz Muehlenhoff wrote:
> > > On Wed, Aug 26, 2009 at 02:25:19PM -0400, Michael S. Gilbert wrote:
> > > > On Wed, 26 Aug 2009 20:01:42 +0200, Moritz Muehlenhoff wrote:
> > > > > On Wed, Aug 26, 2009 at 01:59:58PM -0400, Michael S. Gilbert wrote:
> > > > > > On Wed, 26 Aug 2009 19:29:10 +0200, Moritz Muehlenhoff wrote:
> > > > > > > You should redirect the TODOs in a file separate from CVE/list, 
> > > > > > 
> > > > > > thanks for looking at this.  i personally think that the cve list is
> > > > > > the best destination.  the reasoning is that cve TODOs are good
> > > > > > indicators of what needs worked on and they are associated to specific
> > > > > > cves.  also, the TODOs show up on the security tracker website and are
> > > > > > used by various scripts.
> > > > > > 
> > > > > > yes, the first update from this script will commit over 400 changes,
> > > > > > but assuming those issues are addressed or marked <not-affected>,
> > > > > > subsequent updates will be much smaller.  the important thing is that
> > > > > > running this script increases awareness that a package that you're
> > > > > > dealing with is embedded elsewhere, and for that to be effective, it
> > > > > > needs to update the cve list.
> > > > > > 
> > > > > > > otherwise it clutters the list too much.
> > > > > > 
> > > > > > if you believe that the current formatting is too cluttered, i am
> > > > > > certainly open to suggestions. off the top of my head, for each
> > > > > > affected cve, i could compact the current one line per embed into one
> > > > > > line total for all embeds in that cve.
> > > > > 
> > > > > Working through the list is mostly a QA issue.
> > > > 
> > > > if you want, i can go through the generated list, triage most of the
> > > > TODOs, and mark them either not-affected, fixed, or unfixed before
> > > > uploading any changes; then only a few TODOs (for perhaps complicated
> > > > issues) will remain.  it may require a significant amount of my time to
> > > > get this done, but i'll do what it takes.
> > > 
> > > Great, all confirmed issues can then be added as TODOs to CVE/list.
> > 
> > btw, my script is already smart enough to exclude fixed embeds; it uses
> > the <unfixed>/<removed>/<unknown>/<itp> tags in embedded-code-copies to
> > determine if an issue is open or not.  
> 
> > so as long as the data in
>              ^^^^^^^^^^^^^^
> > embedded-code-copies is accurate, then my script already only generates
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> 
> > TODOs for real issues.  hence, when i am done with this triage, you can
> > expect a lot of new <unfixed> entries (probably close to 200).
> 
> That's exactly why the unfixed/TODO items should _not_ be commited until
> they've been verified manually. There's too much noise in embedded
> code copies (like packages only embedding a small subset of code or
> outdated entries).

understood, and that's why i am going to do this manual triage.  we
should probably force ourselves to be better at keeping
embedded-code-copies up to date and accurate.

mike