silence cast-align warnings from clang on powerpc64

[alfredo]

Summary

silence the following warning when compiling libthr with clang 8 for powerpc64 architecture:

usr/src/lib/libthr/arch/powerpc/include/pthread_md.h:82:10: error: cast from 'uint8_t *' (aka 'unsigned char *') to 'struct tcb *'

increases required alignment from 1 to 8 [-Werror,-Wcast-align]
  return ((struct tcb *)(_tp - TP_OFFSET));

Test Plan

build base, then chroot to new sysroot and run:

tls_test.c737 BDownload

on powerpc64

[emaste]

lib/libthr/Makefile

20–22

I think it would be good to also include a comment about why we know the alignment is not a practical issue.

[kib]

So it is only in _tcb_get() ? Can you use #pragma to only disable the warning at the place ? https://clang.llvm.org/docs/UsersManual.html#controlling-diagnostics-via-pragmas
I do not like too broad settings which might prevent some useful warning ignored.

[bdragon]

In D18807#401468, @kib wrote:

So it is only in _tcb_get() ? Can you use #pragma to only disable the warning at the place ? https://clang.llvm.org/docs/UsersManual.html#controlling-diagnostics-via-pragmas
I do not like too broad settings which might prevent some useful warning ignored.

I agree, it would be nice to have it limited to the actual cast. Combined with perhaps asserting the alignment of the result manually.

[bdragon]

C question: Would a cast through void* be a better workaround for this specific issue by the way?

[alfredo]

In D18807#401486, @git_bdragon.rtk0.net wrote:

In D18807#401468, @kib wrote:

So it is only in _tcb_get() ? Can you use #pragma to only disable the warning at the place ? https://clang.llvm.org/docs/UsersManual.html#controlling-diagnostics-via-pragmas
I do not like too broad settings which might prevent some useful warning ignored.

I agree, it would be nice to have it limited to the actual cast. Combined with perhaps asserting the alignment of the result manually.

Can we assume it's safe to ignore it in the .c is safe for all platforms in any compiler? (actually just for any powerpc* for this copy of pthreads_md.h)

There's another "broader platform" warning that apears if I remove "CFLAGS.thr_stack.c+= -Wno-cast-align" from this Makefile

/usr/src/lib/libthr/thread/thr_stack.c:303:17: error: cast from 'char *' to 'struct stack *' increases required alignment from 1 to 8

[-Werror,-Wcast-align]
          spare_stack = (struct stack *)

lib/libthr/Makefile

20–22

Actually I'm not sure if it's not an issue. I was assuming it's not because this Makefile already contains the following lines:

NO_WCAST_ALIGN.gcc=1 # for gcc 4.2
CFLAGS.thr_stack.c+= -Wno-cast-align

[kib]

In D18807#401490, @git_bdragon.rtk0.net wrote:

C question: Would a cast through void* be a better workaround for this specific issue by the way?

Might be if it works, but I am not sure what compiler authors think about void * alignment requirements. So any way to silence the warning, but make it local to this place.

In D18807#401500, @alfredo.junior_eldorado.org.br wrote:

In D18807#401486, @git_bdragon.rtk0.net wrote:

In D18807#401468, @kib wrote:

So it is only in _tcb_get() ? Can you use #pragma to only disable the warning at the place ? https://clang.llvm.org/docs/UsersManual.html#controlling-diagnostics-via-pragmas
I do not like too broad settings which might prevent some useful warning ignored.

I agree, it would be nice to have it limited to the actual cast. Combined with perhaps asserting the alignment of the result manually.

Can we assume it's safe to ignore it in the .c is safe for all platforms in any compiler? (actually just for any powerpc* for this copy of pthreads_md.h)

There are pragma push/pop ops which should allow you to localize the ignore made with pragmas.

There's another "broader platform" warning that apears if I remove "CFLAGS.thr_stack.c+= -Wno-cast-align" from this Makefile

/usr/src/lib/libthr/thread/thr_stack.c:303:17: error: cast from 'char *' to 'struct stack *' increases required alignment from 1 to 8

[-Werror,-Wcast-align]
          spare_stack = (struct stack *)

It is of similar nature, I believe: the correctly aligned pointer is stored in char *.

[alfredo]

dropped libthr Makefile change and adjusted __tcb_get asm to decrement pointer instead

[alfredo]

fix coding style

[alfredo]

The new version of the patch isn't ignoring or working around compiler warning, it does pointer arithmetic in asm instead, based on idea from git_bdragon.rtk0.net (thanks!).

Tested on PowerPC64 using attached test code. (I cannot test on 32 bit at this moment)

[bdragon]

32 bit version works fine on G4.

[alfredo]

move the "-" signal to input value making it easyer to read

[luporl]

Approving the minor '-' signal move to unblock this change.