Introduce helpers to access pt_regs in an arch-independent manner

[rnav]

This addresses some of the review comments on the RFC patch #522 from @goldshtn, @drzaeus77 and @aaronpuchert.

The additional white-space changes in examples/lua/bashreadline.c are for satisfying style-check.

This should help address #225 and #336. Most of the tests/tools now work on powerpc except for those affected by #525.

[drzaeus77]

Thanks!

[drzaeus77]

FYI, in the future, it is possible (and preferable) to keep the same PR open for multiple iterations of the review cycle. As long as you keep the same branch open, you can keep pushing changes (force push as well) until the reviewers are happy. Keeping it in one PR preserves the comment chain a little more cleanly :)

[rnav]

Thanks - I had been wondering about that. I thought the previous review comments on the patches will be lost if I do a force-push. Your explanation helps.

[4ast]

Seeing errors:
sudo ./funccount.py 'vfs_*'
/virtual/main.c:15:14: warning: implicit declaration of function 'PT_REGS_IP' is invalid in C99 [-Wimplicit-function-declaration]
key.ip = PT_REGS_IP(ctx) & ~3ull;
^
1 warning generated.
LLVM ERROR: Program used external function 'PT_REGS_IP' which could not be resolved!

[rnav]

On 2016/05/07 07:01PM, 4ast wrote:

Seeing errors:
sudo ./funccount.py 'vfs_*'
/virtual/main.c:15:14: warning: implicit declaration of function 'PT_REGS_IP' is invalid in C99 [-Wimplicit-function-declaration]
key.ip = PT_REGS_IP(ctx) & ~3ull;
^
1 warning generated.
LLVM ERROR: Program used external function 'PT_REGS_IP' which could not be resolved!

Weird - I don't see this with my setup on x86/ppc:

/usr/share/bcc/tools/funccount 'vfs_*'

Tracing 47 functions for "vfs_*"... Hit Ctrl-C to end.
^C
ADDR FUNC COUNT
ffffffff812225a0 vfs_fstatat 4
ffffffff81222540 vfs_fstat 5
ffffffff81222440 vfs_getattr_nosec 9
ffffffff81222510 vfs_getattr 9
ffffffff8121bd80 vfs_open 91
ffffffff8121d5c0 vfs_write 97
ffffffff8121d490 vfs_read 264
Detaching...

It looks like clang is not including export/helpers.h for some
reason. If I'm not mistaken, that header should be included
automatically, per commit c597c29.

[4ast]

thanks. false alarm. turned out to be typical double libbcc.so problem.
I did 'rm CMakeCache.txt' which suppose to clear cache only, but it lost the settings of -DCMAKE_INSTALL_PREFIX=/usr and defaulted to /usr/local instead.