On Thursday, June 15, 2017, Tom Lane <
tgl@sss.pgh.pa.us> wrote:
Michael Paquier <michael.paquier@gmail.com> writes:
> On Fri, Jun 16, 2017 at 1:01 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> (2) My inclination would be not to back-patch. This change could break
>> configurations that worked before, and the lack of prior complaints
>> says that not many people are having a problem with it.
> That's fourteen years without complains, still I cannot imagine any
> cases where it would be a problem as people who would have faced this
> problem but not reported it have likely just enforced the FS to handle
> case-insensitivity for paths.
Well, it's not just about case sensitivity. The comment for
SplitDirectoriesString points out that it deals with embedded spaces
differently, and it also applies canonicalize_path(). I'm too tired
to think hard about what that part might mean for compatibility, but
it probably isn't nothing.
Anyway, I agree that this is an appropriate change for HEAD. Just
not convinced that we should shove it into minor releases.
If it's downcasing then careless use of mixed case when the true path on a case sensitive filesystem is all lower case (not unlikely at all given default packaging for deb and rpm packages, iirc) would indeed be a problem. Those paths are accidentally working right now. I vote for no back-patch.
David J.