On Sun, Jun 12, 2022 at 09:22:55PM +0200, Andy Shevchenko wrote:
> > Note there is another patch in this series, which fixes the warning
> > in another way.
>
> > Fixes: fa1451374ebf ("media: atomisp: don't pass a pointer to a local variable")
>
> Dunno for media subsystem, but for ones that Greg is maintain, the
> point is that revert itself is already kinda fix and no need to have a
> Fixes tag, instead the commit message should clearly have the
> automatically generated line of revert (with the rest of the
> explanation why that is needed). Just sharing my experience.
How would that work in this case? We don't have a reference to the git
hash.
The `git revert` command came from early days of git and I always
feel like it hasn't keep up with how git is used these days. The
subject doesn't have the subsystem prefix. The commit message is wrong.
It uses the full git hash instead of the 12 char hash. It doesn't have
a fixes tag. Hans's commit is only correct because he re-wrote
basically everything.
Do a `git --grep=revert`. Some of them you can grep for "This reverts
commit 8bdc2a190105e862dfe7a4033f2fd385b7e58ae8." but there are a lot
which are not machine parsable like:
bd06db5ff9af ("lib/flex_proportions.c: remove local_irq_ops in fprop_new_period()")
4af2bd190a5b ("Revert "squashfs: provide backing_dev_info in order to disable read-ahead"")
646728dff254 ("dmaengine: Revert "dmaengine: add verification of DMA_INTERRUPT capability for dmatest"")
I feel like we should encourage people to not use git revert because
otherwise we're kind of setting them up for failure.
regards,
dan carpenter
