On Thu, 2026-05-28 at 19:19 +0000, Timur Tabi wrote: > On Thu, 2026-05-28 at 15:09 -0400, [email protected] wrote: > > You are right - I did actually sit down and try to run through all > > of > > the codepaths for all these functions, but there's clearly a number > > that I missed and should have checked sashiko for. > > Even if you could narrow down the list to those functions that are > 100% okay with IS_ERR, that's not > future proof. I wouldn't be surprised if there are some code paths > where IS_ERR_OR_NULL is > required, but then the NULL is converted to an ERR, and so the top- > level caller will see only ERR. > But what happens if there's a refactor or some other major code > change? > > Like I implied earlier, the real problem is that Nouveau is > inconsistent in how it returns failure. > Sometimes it returns NULL, sometimes it returns ERR, sometimes it > could be either. So a proper fix > is not replacing IS_ERR_OR_NULL with IS_ERR wherever we can get away > with it today. The proper fix > is to revamp the entire code base to be consistent, and no one wants > to do that work.
Yeah - that's a good point. I will go ahead and send the revert for the whole series then, would you mind helping to review it?
