fix: use Padding in stat padding fields

[xtqqczze]

Use Padding in stat and stat64 for fields like st_spare, __pad, __unused

[tgross35]

I think this may overlap with #5444

[dybucc]

This change is not part of #5444.

View changes since the review

[dybucc]

This change is not part of #5444.

View changes since the review

[dybucc]

This change is not part of #5444.

View changes since the review

[dybucc]

This change is not part of #5444.

View changes since the review

[dybucc]

This change is not part of #5444.

View changes since the review

[dybucc]

I just looked through the patch, and left some comments where there were
changes not included in #5444. They are minimal, but some of them affect
previously public fields. I can include them in #5444, but that is not
for me to decide.

[xtqqczze]

@dybucc I don't think you should include the additional changes in #5444. If necessary the changes such as removing pub from __st_ino, __old_st_blksize and __old_st-* could be made in a seperate PR.

[dybucc]

@dybucc I don't think you should include the additional changes in
#5444. If necessary the changes
such as removing pub from __st_ino, __old_st_blksize and
__old_st-* could be made in a seperate PR.

Sure thing.

[xtqqczze]

I’ve removed the unnecessary field-visibility changes from this PR, so it is now superseded by #5444.

However, I note that #5444 was declined for stable. These changes are necessary to prevent Undefined Behavior, as reported in bytecodealliance/rustix#1668.

@tgross35 Since this PR is already targeted at libc-0.2, could these changes still be considered for inclusion here?

[tgross35]

Realizing now this is targeted for 0.2. I won't want to take this change since these are some of the most popular structs in the crate and they happen to have all fields public on some platforms. Just means there's a higher chance of breakage than most other changes like this.

View changes since this review

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[tgross35]

However, I note that #5444 was declined for stable. These changes are necessary to prevent Undefined Behavior, as reported in bytecodealliance/rustix#1668.

Also, commented on the other thread but so it's here as well: libc's definitions are correct, it's not our bug. rustix is doing a bad assume_init, among many others, and Miri just happened to flag the problem on stat.