On Wed, Mar 25, 2026 at 11:57:38AM -0600, Shuah Khan wrote:
> My primary concern with this change is that it 10 files changed, with
> 26 insertions(+), 32 deletions(-)
> 
> It can be difficult to find any regressions unless the changes are
> tested. I understand it is staging repo, but if we relax the rules
> for staging irrespective of the scope of change, it becomes lot
> harder to find regressions later. We are essentially leaving the
> job of testing to users. Also if ans when the driver is pulled into
> mainline, it will inherit the regressions that crept in.
> 

To me this is a mechanical patch (basically automatic, flip the
if statement) from a newbie.  These kinds of patches are seldom a
problem.

I took a sample of the patches for five years from 2021 to the end of
2025.  We merged 54406 commits in staging.  We had 401 Fixes tags.  But
then I decided I don't care about Kconfig problems so I cut it down to
380 bugs.  Half of those (189 commits) were from when the driver was
merged.  We tend to not review code very much when it is first merged
to staging.  The other half (191 commits) were from us missing bugs
during review.  191 bugs over five years is 40 bugs per year.

I hand reviewed a bunch of the 191 commits.  It's mostly senior
developers and maintainers who introduce bugs.  They are normally doing
complicated things like adding features or re-working core parts of the
driver.  Quite a few of these patches were tested.

The fallout from newbie patches is pretty rare and often really
minor.  A common thing is "We deleted code but we didn't delete
everything."  The number of serious mess ups is probably just a
couple times per year out of the 40 total bugs per year.

I've attached my script so you can check yourself.  NEW means the
bug was introduced by a new driver.  OLD means, ideally it would
have been caught in review.

regards,
dan carpenter