Messages in this thread | | | Subject | Re: [RESEND PATCH v5 3/5] drivers core: allow probe_err accept integer and pointer types | From | Andrzej Hajda <> | Date | Wed, 24 Jun 2020 21:39:57 +0200 |
| |
On 24.06.2020 17:16, Robin Murphy wrote: > On 2020-06-24 16:04, Mark Brown wrote: >> On Wed, Jun 24, 2020 at 03:25:33PM +0100, Robin Murphy wrote: >> >>> And yeah, anyone who pipes up suggesting that places where an >>> ERR_PTR value >>> could be passed to probe_err() could simply refactor IS_ERR() checks >>> with >>> more uses of the god-awful PTR_ERR_OR_ZERO() obfuscator gets a long >>> stare of >>> disapproval... >> >> We could also have a probe_err_ptr() or something that took an ERR_PTR() >> instead if there really were an issue with explicitly doing this. > > Yeah, for all my lyrical objection, a static inline <blah>_ptr_err() > helper to wrap <blah>_err() with sensible type checking might actually > be an OK compromise if people really feel strongly for having that > utility.
I have proposed such thing in my previous iteration[1], except it was macro because of variadic arguments.
With current version we save 8 chars and hacky macro, with the old version we save only 4 chars and more clear construct - less tempting solution for me.
Personally I prefer the current version - it does not seems to me more dangerous than all these PTR_ERR, IS_ERR,ERR_PTR helpers, but can prevent expression split across multiple lines due to 80char limit.
Probably the simplest solution is to drop this patch, I will do it then.
[1]: https://lwn.net/ml/linux-kernel/20181220102247.4911-4-a.hajda@samsung.com/
Regards
Andrzej
> > (and then we can debate whether it should also convert NULL to -ENOMEM > and !IS_ERR to 0... :D) > > Robin. > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > https://protect2.fireeye.com/url?k=074420c0-5ada8e5a-0745ab8f-0cc47a336fae-bba8bb4caf96e14d&q=1&u=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Fdri-devel > >
| |