I was not very happy about the old naming for the macros so I have new
names which I think are more descriptive:

Just IMHO, and not meaning to stand in the way ...

Trying the _Pragma suddenly doesn't look like a bad idea after all: if GCC adopts
the attributes it would be easier to adapt them in just one place instead of chasing
all the headers.

In D9007#189888, @email_piotr-stefaniak.me wrote:

After learning about PR 215690 I'm no longer convinced that the regression tests I'd like to have should be integrated into the FreeBSD testing suite. It sounds ridiculous that a harmless regression in a reformatting tool like indent(1) would create "bogus failures counted against our overall results". indent(1) is not a crucial part of the operating system and shouldn't be tested as rigorously. Bugs in indent(1) are not even close to an indication of overall quality of the system.

In D9004#189482, @pfg wrote:

In D9004#189480, @ed wrote:

I like the idea, but I think it would make more sense to do this a bit more slowly. For example, we could use this change to add stubs for these keywords for old compilers to sys/cdefs.h.

Add check for nullability feature before using the pragmas.
We still have many older versions of clang in the ports tree.

Looking at deprecating the __nonnull() attribute completely, I found a
couple of occurrences in lib/libthr/thread/thr_private.h.

Update copyinstr().

In D9004#188501, @kib wrote:

In D9004#188458, @pfg wrote:

Update to bring the pop pragmas.

I left the copyinstr(9) as it is for consistency with the previous
declaration. FWIW, according to the manpage the first argument should be a
"NUL-terminated string", not sure conceptually a NULL pointer would
classify.

0 is absolutely (potentially) valid pointer to a userspace string. More, in past there were ABIs where there was a zero-length string located at address zero.

Update to bring the pop pragmas.

In D9004#188426, @kib wrote:

Why is this useful at all ? IMO it adds enormous and useless noise to the headers, while providing a doubtful value, if any.

Note that I am only replacing the very dangerous GCC __nonnull__ attribute with the much more benign clang _Nonnull. Most cases were pre-existing, except for some functions where we are followinh Apple Darwin, in the hope of being more firendly to swift (and clang in the Mac)

Add kib per the pthread changes.

Update some of the threading functions.
Move the pragmas consistently to the top of the header.
This is a commit candidate.

Nice catch, Coverity missed it completely.

In D9007#186867, @ngie wrote:

In D9007#186865, @pfg wrote:

...

I am not sure about the MFC: There are 30-something changes in
head that have not been MFC'd (and I am just too lazy to merge them).

:{..

Code should be MFCed though if possible -- otherwise you won't get feedback on your commits until the next .0 release cycle (which will be a couple years from now).

In D9007#186863, @ngie wrote:

In D9007#186862, @pfg wrote:

In D9007#186861, @email_piotr-stefaniak.me wrote:

In D9007#186860, @ngie wrote:

I have a potential compromise: please remove the "Makefile" from the commit, integrate your tests into the FreeBSD test suite separately, open another CR for that change, and I'll approve that change once it's complete.

Do you mean it'd be OK to commit the test cases (*.c, *.good, *.pro) without the Makefile? I'd be fine with that.

This doesn't make much sense to me but as I said, if ngie@ approves it I am OK with it too. Please do state in the commit log that the tests are currently disconnected from the build but that you intend to complete them. As before, rember to include a link to the CR.

I can help guide piotr with the integration work, once the test content is present.

In D9007#186861, @email_piotr-stefaniak.me wrote:

In D9007#186860, @ngie wrote:

I have a potential compromise: please remove the "Makefile" from the commit, integrate your tests into the FreeBSD test suite separately, open another CR for that change, and I'll approve that change once it's complete.

Do you mean it'd be OK to commit the test cases (*.c, *.good, *.pro) without the Makefile? I'd be fine with that.

In D9007#186751, @ngie wrote:

...
Put differently, I cannot approve this because it doesn't really integrate into the existing system I contribute to (the FreeBSD Test Suite). If pfg@ approved the commit, it's fine to commit as-is.

In D9007#186525, @pfg wrote:

...

ngie is listed as MAINTAINER for tests, so in this case approval by ngie means I approve it too.

In D9004#186600, @emaste wrote:

I see, but you write

Turning off -Wnullability-completeness is not an option.

I'm curious why it's not an option.

In D9004#186583, @emaste wrote:

I suspect this is fine (haven't yet looked in detail). Is turning off -Wnullability-completeness a nonstarter because we have some headers that are already completely annotated and want to prevent new, unannotated declarations from appearing?

More cleanups: the pragmas have to appear before *any* pointer is
declared.

Having a license is important and the committer guide specifies a two clause BSD license as the preferred. for new files In this case the license uses more space that the tests though, so perhaps it is worth discussing in the lists the possibility of using a shorter statement like CC0. IANAL ... so I generally avoid such discussions.

Approved: however, do remember to add a line

Update with small cleanups that were breaking world.

"-P" is OK then.
I can't check easily the length of the lines in phabricator but it's probably fine too as I see longer lines in the original code.

Given that there is a "-npro" option already, how about renaming it to "-ppro"? This would stand for "personalized profile", but I guess we could use custom profile (-cpro) or temporary profile (-tpro). Just trying to make things look consistent ;).

I figured it out:

The '__nonnull()' attribute is rather dangerous, it seems reasonable to replace it with the '_Nonnull' qualifier. Unfortunately it seems like a lot more changes than those from either bionic or Apple are required to build cleanly. Turning off -Wnullability-completeness is not an option.

Use of these attributes breaks the build in ugly ways. For example:

Thanks for the review: we do have old tests lying around that don't integrate with ATF. I guess it's sensible to expect that all new test do so. Thanks for the pointers.

I am glad you are passing this through phabricator. I don't have much experience with the testing framework. Let me add the local expert.

Minor updates.
While here I should mention, I did strip a bit the indentation in
pthread.h to make it more readable .

Learn a bit more from Bionic's libc.

In D3189#182628, @adrian wrote:

so it's just missing those atomic ops for MIPS32/MIPS64? It compiles/works fine on other architectures?

This really needs more help.

Bring some fixes from NextBSD.