Re: [GIT PULL] iomap: new code for 5.13-rc1

[Rasmus Villemoes]

On 28/04/2021 18.50, Linus Torvalds wrote:
> [ Added Andy, who replied to the separate thread where Jia already
> posted the patch ]
> 
> On Wed, Apr 28, 2021 at 12:38 AM Rasmus Villemoes
> <linux@xxxxxxxxxxxxxxxxxx> wrote:
>>
>> So the patch makes sense to me. If somebody says '%pD5', it would get
>> capped at 4 instead of being forced down to 1. But note that while that
>> grep only produces ~36 hits, it also affects %pd, of which there are
>> ~200 without a 2-4 following (including some vsprintf test cases that
>> would break). So I think one would first have to explicitly support '1',
>> switch over some users by adding that 1 in their format string
>> (test_vsprintf in particular), then flip the default for 'no digit
>> following %p[dD]'.
> 
> Yeah, and the "show one name" actually makes sense for "%pd", because
> that's about the *dentry*.
> 
> A dentry has a parent, yes, but at the same time, a dentry really does
> inherently have "one name" (and given just the dentry pointers, you
> can't show mount-related parenthood, so in many ways the "show just
> one name" makes sense for "%pd" in ways it doesn't necessarily for
> "%pD"). But while a dentry arguably has that "one primary component",
> a _file_ is certainly not exclusively about that last component.
> 
> So you're right - my "how about something like this" patch is too
> simplistic. The default number of components to show should be about
> whether it's %pd or %pD.

Well, keeping the default at 1 for %pd would certainly simplify things
as there are much fewer %pD instances.

> That also does explain the arguably odd %pD defaults: %pd came first,
> and then %pD came afterwards.

Eh? 4b6ccca701ef5977d0ffbc2c932430dea88b38b6 added them both at the same
time.

Rasmus