Actually, scratch that, I can confirm that Android has their permissions
set to 711 on the /data folder. Otherwise we wouldn't be able to cd to
them. And the 1 bit *is* the searchable bit (not the executable bit) for
directories (TIL), so /data is searchable.

So, both the pre-patch and post-patch behaviors are noncompliant. The first
is too strict and this is too lenient (although I prefer more lenient than
strict). Back to the drawing board, I guess.

On Tue, Jan 12, 2021, 7:53 PM Devin Hussey <husseydevin@gmail.com> wrote:

> We are already violating POSIX with our globbing:
>
> "If the pattern does not match any existing filenames or pathnames, the
> pattern string shall be left unchanged."
>
> Therefore, this:
>
> zsh:1: no matches found: /tmp/inaccessible/a/*
>
> is wrong, it should just print this:
>
> /tmp/inaccessible/a/*
>
> Also, at least with the implementation on my device, the glob() function
> works fine as long as the folder has execute permission. (The sample was
> wrong, it should be 111, not 000)
>
> On Tue, Jan 12, 2021, 6:47 PM Lawrence Velázquez <vq@larryv.me> wrote:
>
>> > On Jan 12, 2021, at 5:42 PM, Devin Hussey <husseydevin@gmail.com>
>> wrote:
>> >
>> > Even if we can't access a parent folder, there is still a chance we can
>> > access a subdirectory.
>>
>> I might be mistaken (entirely possible!), but this behavior seems
>> to violate POSIX volume 3 chapter 2 section 2.13.3, which states
>> in part that
>>
>>         Specified patterns shall be matched against existing filenames
>>         and pathnames, as appropriate. Each component that contains a
>>         pattern character shall require read permission in the directory
>>         containing that component. Any component, except the last, that
>>         does not contain a pattern character shall require search
>> permission.
>>
>> (
>> https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_13_03
>> )
>>
>> vq
>
>