Am 07.08.25 um 17:06 schrieb Dan Carpenter:
On Thu, Aug 07, 2025 at 04:27:41PM +0200, Stefan Metzmacher wrote:
The DB is too big and too dependent on your .config but I should
share the smatch_data/ more regularly. I started to push that into
a separate git repo but I didn't finish that work. I should do
that.
Ok, what's the gain of updating it?
Does it help when doing fixes on old kernels?
If you run smatch_scripts/build_kernel_data.sh then it runs
smatch_scripts/gen_* which generates a bunch of the files in
smatch_data/. Which in theory should enable more warnings for new
code.
I've been moving away from generating files and more towards
putting everything in the DB. I just took a look at the files now
to respond to your email and what I saw wasn't good... I need to
look at this some more.
I don't know how often the zero day bot rebuilds the smatch_data.
I bet they never do and so I think it doesn't really matter.
I'm typically doing a full kernel build a week after each rc.
My idea was to rebuild the whole db after doing that.
Yeah. That's a good strategy. The data from the existing DB feeds
into the new one when you rebuild the DB so don't delete the old
DB at the start or anything.
I mean I'm typically do a git clean -xdf . in order
to wipe everything in order to do a clean:
make -j33 bindeb-pkg
So it would build a new DB, as I'm working based on
the new kernel I guess that's all I need or
are there other reasons to update the existing DB?
Thanks!
metze
