This works similarly to how printf'ing a Type::usym works.

When verifying the printf format string we treat stack and ustack as
Type::string. BPFtrace:get_arg_values now knows how to resolve these.

Fixes #10

This makes cgroupid function to work fine when building with alpine.

Allow '.' in idents for golang

Supports:
- LLVM 5 (static build)
- LLVM 5
- LLVM 6
- LLVM 7

ClangParser: Don't exit if kernel headers are not found

When specifying a string literal attach point do not try to expand
wildcards if they are present.

This specifically is required for Go methods on pointer receivers,
which have a '*' in their symbol name.

Allow quoted string attach points so that you can match more than
[a-Z0-9].

This specific issue that this fixes is attaching to go
functions. These symbols are in the format '<package>.<function>' and
can contain unicode letters and digits.

Implement function cgroupid(path) -> cgroup id

Mention the `cgroup` variable and the `cgroupid` function alongside an
example that uses them both.

Currently used image of alpine has a version of musl libc that does
not have name_to_handle_at function that is required to resolve the
cgroup id. In such case we just use an implementation that says that
resolving cgroupid is not supported.

We will wrap the file_handle struct in the next commit. The struct has
a flexible array member, which is not supported by C++. Compiler may
complain about using it when allocated on stack, even indirectly as a
member of a struct. I'm not sure if using this kind of types is even a
defined behavior…

Initial LLVM 7 port

Apply Brendan's patch

* fixes: https://github.com/iovisor/bpftrace/issues/9

* test fix

* tests

* lifetime

* crash fix

Extend support for char type

Fix link to count function in reference guide

This fixes the link to the count() function under the Frequency
Counting section.
Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>

This extends support for the char type by making the clang
parser recognize CXType_Char_U as well.

Currently, the clang parser assumes that the char type is
always implemented as a signed char. However, for certain
architectures like powerpc64, the char type is unsigned
and the parser fails to recognize such variables.
Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>

If make is not installed, a clean Fedora install fails with:

CMake Error: CMake was unable to find a build program corresponding to "Unix
Makefiles".  CMAKE_MAKE_PROGRAM is not set.  You probably need to select a
different build tool.

Advise users to install make to resolve this.
Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com>

- KBuildHelper takes kdir, not kpath, as a parameter.
- Copy BCC and set has_kpath_source to false when reading
  kpath from environment variable

BPFTRACE_KERNEL_SOURCE  environment variable

* add bpftrace man page

* changes to bpftrace man page

* fix bpftrace manpage file name

Still continue regular clang parsing, just without extra kernel header flags

The previous fix was bad as it elided a correct early return.

Supress invalid file warning for wildcarded tracepoints

Fix name of kernel required config option in INSTALL.md

Wildcards, unless expanded, are not valid tracepoints. We should not
warn about missing files if the patch is wildcarded. Interestingly enough,
the previous behavior (early return) could have caused fewer tracepoints
to be attached than available. However in practice, this doesn't seem
to happen.

This closes issue #133.

In Linux Kernel 4.7, CONFIG_HAVE_BPF_JIT was splitted into two
(CONFIG_HAVE_EBPF_JIT and CONFIG_HAVE_CBPF_JIT) for distinguishing cBPF
and eBPF JITs
(https://www.mail-archive.com/netdev@vger.kernel.org/msg110538.html)

Fix typos

Add filter by cgroup-v2 id

INSTALL.md: add clang-devel as Fedora build dep

To get the cgroup id, you can use the following program to call
name_to_handle_at(), borrowed from
linux/tools/testing/selftests/bpf/cgroup_helpers.c:
```
wget https://gist.githubusercontent.com/alban/4a9aea9636d9fdf11709eddd21df4cc4/raw/f4fc915483a069d6044957e27df091923f95c48b/getcgroupid.c
gcc -Wall -o getcgroupid getcgroupid.c
./getcgroupid /sys/fs/cgroup/unified/
0x100000001
4294967297
```

To test it, I run the tracer in one terminal:
```
./build.sh
sudo ./build-release/src/bpftrace -e 'tracepoint:syscalls:sys_enter_openat /cgroup == 0x100000001/ { printf("%s\n", str(args->filename)); }'
```

And some tests in another terminal:
```
sudo -s
echo $$ > /sys/fs/cgroup/unified/cgroup.procs
cat /dev/null
```

And observe the output of the tracer:
```
Attaching 1 probe...
/etc/ld.so.cache
/lib64/libc.so.6
/usr/lib/locale/locale-archive
/dev/null
^C
```