Add .gitignore entry for .theia

This directory is used for project-specific settings for the Theia IDE.
Since I use that, I would like if the directory was ignored.  Even
though it doesn't apply to everybody, it shouldn't bother anybody to
have this entry.  Also, I wouldn't mind if we added similar entries for
other IDEs, if it helped other people.

Change-Id: Ic6f42b9d3f1d855e3d6e9e0dee1492aa484010ac
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1327
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: logging: possible buffer overflows

Issue
=====
Multiple possible buffer overflows on string operations using the
`SET_TMP_PREFIX()` macro that uses `strcpy()` on parameter char array.

Solution
========
Use a #define to set the length of the destination array and use it as
the size parameter of the `strncpy()` and `strncat()` calls.

Drawbacks
=========
None.

Notes
=====

Coverity reported defects:
  CID 1401179 (#5 of 5): Copy into fixed size buffer (STRING_OVERFLOW) 9.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401181 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 5.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length

  CID 1401186 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 15.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401192 (#2 of 2): Copy into fixed size buffer (STRING_OVERFLOW) 4.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401197 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 6.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401198 (#4 of 4): Copy into fixed size buffer (STRING_OVERFLOW) 15.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401203 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 28.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401212 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 12.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401215 (#4 of 4): Copy into fixed size buffer (STRING_OVERFLOW) 16.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401221 (#2 of 2): Copy into fixed size buffer (STRING_OVERFLOW) 19.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401227 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 13.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401230 (#3 of 3): Copy into fixed size buffer (STRING_OVERFLOW) 10.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401232 (#3 of 3): Copy into fixed size buffer (STRING_OVERFLOW) 23.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401234 (#5 of 5): Copy into fixed size buffer (STRING_OVERFLOW) 19.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401254 (#10 of 10): Copy into fixed size buffer (STRING_OVERFLOW) 10.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401257 (#1 of 1): Copy into fixed size buffer (STRING_OVERFLOW) 9.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

  CID 1401261 (#2 of 2): Copy into fixed size buffer (STRING_OVERFLOW) 11.
  fixed_size_dest: You might overrun the 64-character fixed-size string
  tmp_prefix by copying prefix without checking the length.

Reported-by: Coverity - Copy into fixed size buffer (STRING_OVERFLOW)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I922f1fb82a95e06b0c42627a2e57ba94debe1c5a
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1329
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins

Add vim .swp files to gitignore

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: Ibfe600455d80a8542a8f7ebcda44f3682b4b0534
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1330
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Replace g_build_path with g_build_filename for portability

g_build_filename will use the correct separator depending on the
platform.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: Iaeeb1a4c30aaea7690ba479ba05108c6fc57dd4e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1331
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: test_bitfield: buffer overrun accesses

This issue is due to a type mismatch between union member vs type passed
as parameter to the macros as found by Coverity:

** CID 1401409:  Memory - corruptions  (OVERRUN)
/tests/lib/test_bitfield.c: 450 in run_test_signed_write()

Issue introduced by commit 0675eb8c07 "Extend test_bitfield coverage"

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Change-Id: I14c9d69011b369e5dc8be4305ac8bf32cc86d3dd
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1321
CI-Build: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: tests: use underscores in filename

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I02a8b61e7c4a234aa017a941f4609f4276f387ee
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1320
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: flt.utils.muxer: avoid returning a pointer needlessly

Reshuffle `muxer_msg_iter_add_upstream_msg_iter()` function to return an
int rather than a pointer.

Returning a pointer in this case has no value since the newly allocated
structure is stored in a GQueue directly. Returning an int to express
success or error seems more appropriate and less error-prone.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I111cd06e1f667c98aa3bf650375c6e49fcd4c5dd
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1313
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: flt.utils.muxer: Potential memory leak

scan-build reports a potential memory leak of the
muxer_upstream_msg_iter pointer allocated in the
`muxer_msg_iter_add_upstream_msg_iter()` function if the `g_queue_new()`
fails.

So, free the muxer_upstream_msg_iter on the error path.

Reported-by: scan-build - Potential leak of memory
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia37a1be9525b357d063d5c5bd5623ddccbd4e29d
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1312
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: bitfields: streamline use of underscores

Do not prefix macro arguments with underscores. Use one leading
underscore as prefix for local variables defined within macros.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Change-Id: Ie1c2f7f59f605ac62d483aba67b3f70cef27bf21
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1318
Tested-by: jenkins

Silence compiler "always false comparison" warning

Compiling the bitfield test with gcc -Wextra generates those warnings:

 ../../include/babeltrace/bitfield-internal.h:38:45: warning: comparison of unsigned expression < 0 is always false [-Wtype-limits]
 #define _bt_is_signed_type(type) ((type) -1 < (type) 0)

This is the intent of the macro. Disable compiler warnings around use of
that macro.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Change-Id: I2dd980e11ebd6cd37ae71b013925a6fca5d7b483
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1307
Tested-by: jenkins

Extend test_bitfield coverage

test_bitfield was mainly testing various write unit size. Add
variations of read unit size as well.

Previously, the test was only covering input from a 32-bit integer.
Additionally test source and destination of 64-bit.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Change-Id: Ic2b7849140e04fe7cca3de576e31bfef8f0a03ae
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1305
Tested-by: jenkins

Fix: bitfield: shift undefined/implementation defined behaviors

bitfield.h uses the left shift operator with a left operand which
may be negative. The C99 standard states that shifting a negative
value is undefined.

When building with -Wshift-negative-value, we get this gcc warning:

In file included from /home/smarchi/src/babeltrace/include/babeltrace/ctfser-internal.h:44:0,
                 from /home/smarchi/src/babeltrace/ctfser/ctfser.c:42:
/home/smarchi/src/babeltrace/include/babeltrace/ctfser-internal.h: In function ‘bt_ctfser_write_unsigned_int’:
/home/smarchi/src/babeltrace/include/babeltrace/bitfield-internal.h:116:24: error: left shift of negative value [-Werror=shift-negative-value]
   mask = ~((~(type) 0) << (__start % ts));  \
                        ^
/home/smarchi/src/babeltrace/include/babeltrace/bitfield-internal.h:222:2: note: in expansion of macro ‘_bt_bitfield_write_le’
  _bt_bitfield_write_le(ptr, type, _start, _length, _v)
  ^~~~~~~~~~~~~~~~~~~~~
/home/smarchi/src/babeltrace/include/babeltrace/ctfser-internal.h:418:3: note: in expansion of macro ‘bt_bitfield_write_le’
   bt_bitfield_write_le(mmap_align_addr(ctfser->base_mma) +
   ^~~~~~~~~~~~~~~~~~~~

This boils down to the fact that the expression ~((uint8_t)0) has type
"signed int", which is used as an operand of the left shift.  This is due
to the integer promotion rules of C99 (6.3.3.1):

    If an int can represent all values of the original type, the value is
    converted to an int; otherwise, it is converted to an unsigned int.
    These are called the integer promotions. All other types are unchanged
    by the integer promotions.

We also need to cast the result explicitly into the left hand
side type to deal with:

warning: large integer implicitly truncated to unsigned type [-Woverflow]

The C99 standard states that a right shift has implementation-defined
behavior when shifting a signed negative value. Add a preprocessor check
that the compiler provides the expected behavior, else provide an
alternative implementation which guarantees the intended behavior.

A preprocessor check is also added to ensure that the compiler
representation for signed values is two's complement, which is expected
by this header.

Document that this header strictly respects the C99 standard, with
the exception of its use of __typeof__.

Adapt use of _bt_piecewise_lshift in plugins/text/pretty/print.c
to the new API.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I47d2655cfe671baf0fc18813883adf7ce11dfd6a
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1311
Tested-by: jenkins

Fix: flt.utils.trimmer: bt_message_put_ref() on freed message

Issue
=====
In this function, we convert a timestamp in second to a clock value in
order to create a stream activity beginning message. We create a new
message and proceed to call `clock_raw_value_from_ns_from_origin()` to
convert the timestamp to a clock value.

If this call fails, we put the reference on this message using
`bt_message_put_ref()`. This frees the message as the refcount was at 1
because it was just created.

The problem here is that we don't reset the `msg` pointer and return it
to the call via the output parameter. This pointer points to freed data.
The caller then figures that the function failed by looking at the
return value and proceed to call `bt_message_put_ref()` again resulting
and assertion failure.

This can be easily triggered by passing a `begin` parameter of value 0
to a trimmer component.

Solution
========
Use a local variable to keep the pointer to the newly created message
and move that reference to the output parameter only on success.

Drawbacks
=========
None.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Id589f25d9ec6f555abd44ccb71e72af8e1684272
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1314
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: flt.utils.muxer: Undefined or garbage value returned

Reported-by: scan-build - Undefined or garbage value returned to caller
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I46c6e431ae4dcb68367dea84971779af017ecded
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1310
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

fix: Add PROT_READ to ctfser mmap flags

On Linux PROT_WRITE implies PROT_READ but on other Unices like Solaris,
it's not the case. This resulted in a segfault when reading from the
mapping.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: If11bbcf32e33bf5e682afaade2d3769a141c6583
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1302
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins

Fix: src.ctf.fs: possible use-after-free in the error path

Reported-by: scan-build - Use of memory after it is freed
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I5c19a45b0ddd11eed3872295efeb7b5201967419
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1303
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix type-limits warnings, uint can't be < 0

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I14ba4492bd2ea9803daf8985dd8fe21107942df6
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1295
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix old-style-declaration warnings

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I997662eaaabc217b22a0010dccd292e4e946933f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1294
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Annotate implicit fall-through

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I21c5641869afca1fcac301f612d02d2cb10583db
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1293
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: src.ctf.fs: index no longer used in loop

This removal should have been done in an earlier commit modifying this
loop:
  Author: Francis Deslauriers <francis.deslauriers@efficios.com>
  Date:   Tue Apr 30 19:01:17 2019 -0400

      src.ctf.fs: compute stream range using entire file info group

      Given that index entries are sorted within their file info structure and
      that file info structures are sorted by time within the file info group,
      we can simply compute the range of the stream by using the absolute
      first index entry and the absolute last index entry of the entire file
      info group.

Add `BT_ASSERT()` to enforce invariants in developer mode.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I4e0b53cbb808835886b97b6d6001ecd7bcc2abb4
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1288
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: sink.ctf.fs: Dereference before null check

No need to check if `trace->metadata_path` is null at the point because
it has been used by `fopen()` already and `abort()` is called if
`fopen()` fails.

  CID 1401238 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking trace->metadata_path suggests that it
  may be null, but it has already been dereferenced on all paths leading
  to the check.

Reported-by: Coverity (1401238) Dereference before null check
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I90590323764810da8a6b4cfb37445a588befd242
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1284
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins

Fix: sink.ctf.fs: possible uses after free because of unchecked return values

Uses after free are possible because we ignore the return value of the
`bt_ctfser_write_*()` function calls. To avoid this, check the return
values and log an error message.

  CID 1401217 (#4 of 4): Use after free (USE_AFTER_FREE)
  11. deref_arg: Calling bt_ctfser_write_byte_aligned_unsigned_int
  dereferences freed pointer stream->ctfser.base_mma

  CID 1401193 (#2 of 2): Unchecked return value (CHECKED_RETURN)
  8. check_return: Calling bt_ctfser_write_byte_aligned_unsigned_int
  without checking return value (as is done elsewhere 8 out of 10
  times).

Reported-by: Coverity (1401217) Use after free
Reported-by: Coverity (1401193) Unchecked return value
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I7581c0e549d7bf916f42afe306976baa9822897a
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1286
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: common/assert.c: remove useless calls

  CID 1401190 (#1 of 1): Useless call (USELESS_CALL)
  side_effect_free: Calling bt_common_color_bold() is only useful for
  its return value, which is ignored.

  side_effect_free: Calling bt_common_color_fg_red() is only useful for
  its return value, which is ignored.

Reported-by: Coverity (1401190) Useless call
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia25823f44c24997782ccc209c48d5b28a45a98f4
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1287
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Cleanup: src.ctf.fs: erroneous comment

The `len` field is already increased by the `g_ptr_array_add()` call.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I6c261aa20ca79b368c668d0e77879611e56b0bc6
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1289
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Add -Werror=implicit-function-declaration to AM_CFLAGS

Implicit declaration of function is most likely always a bad thing.

Implicit declaration of function is particularly problematic with
Babeltrace 2 because of its plugin system. Currently, when building a
plugin shared object, the compiler issues a warning when it encounters
an implicit declaration but will create the .so just fine but with the
implicit function marked as undefined, as expected by the standard prior
 to C99.

When trying to load a Babeltrace plugin with an undefined symbol, glib's
`g_module_open()` fails and Babeltrace then considers that this .so is
not a valid Babeltrace plugin.

A failure to load a plugin is not a error because we want Babeltrace to
be able to list/use every plugins in a given directory by looking at
every .so file. So we can't simply return an error in those cases.

Since Babeltrace marks plugins with undefined symbols as invalid they
can't be used in a graph.

This problem is caused when a typo in a function name finds its way into
the code and a warning message is ignored. It can also happen when the
code uses a function of more recent library than what is currently
installed on the system.

So, to ensure this never happens, we promote the
`-Wimplicit-function-declaration` warning to the rank of compilation
error.

This is diagnostic has been supported by GCC since v4.8 and Clang since
v4 at least so we believe it's safe to enable it by default.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia43b4f8c85b6a6cb7557f084a2bed07e7ec043dc
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1278
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Add missing gitignore for test_ctf_plugin

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I903f6c119ec51e5c96b555f762bcd0d43e11446f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1285
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Harmonize pprint macro across projects

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I9d02bf7e3c438162c29ed58b1ca2050b76d15363
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1274
CI-Build: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Update macros from the autoconf archive

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I6d496de7d988c47ebeaf314e9639281853116829
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1273
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

flt.lttng-utils.debug-info: fd-cache: log to `debug` severity on stat() error

Issue
=====
A `debug-info` component tries to open multiple files in its search to
find the debugging information necessary to resolve the addresses
contained in the trace.

Early in the `bt_fd_cache_get_handle()` function, a `stat()`
is done on the path to get the inode number and device number. This
`stat()` returns an error if the file is absent. Currently, in those
cases, an message is logged at the `error` severity level (BT_LOGE_*)
resulting in the printing of an error message. Since multiple files are
tired until the right one is found (if any), the user can see multiple
error messages even if the right file may be found later.

This is undesirable because it mislead the user into thinking that an
error occurred when in fact it's completely normal.

Solution
========
Log the failure of this `stat()` call to the `debug` severity level.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: If81f4333374e6ee95b4dea07924c3d37f0e3b652
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1276
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: src.ctf.fs: Using uninitialized value

`iter_status` may be used uninitialized if the break statement in the
do-while is reached.

Fixes Coverity #1401236:
  CID 1401236 (#1 of 1): Uninitialized scalar variable (UNINIT)
  9. uninit_use: Using uninitialized value iter_status.

Reported-by: Coverity (1401236) Uninitialized scalar variable
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I96e4608011144cb3c1ae26018d57e98f0358a779
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1261
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: sink.ctf.fs: writing 64bit real number as 32bit

Both 32bit and 64bit Trace IR real numbers are written as 32bit CTF
floats with the `bt_ctfser_write_float32()`. This is most probably a
copy-paste error.

To fix this, use the `bt_ctfser_write_float64()` in the else branch instead.

Coverity report:
  CID 1401207 (#1 of 1): Identical code for different branches (IDENTICAL_BRANCHES)
  identical_branches: The same code is executed regardless of whether
  fc->base.size == 32U is true, because the 'then' and 'else' branches
  are identical. Should one of the branches be modified, or the entire
  'if' statement replaced?

Reported-by: Coverity (1401207) Identical code for different branches
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I462fd47e968e39f5878431d5ac6cbcbbf1d2ffcf
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1268
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: flt.lttng-utils.debug-info: `ip` field is 32bit on 32bit cpus

Issue
=====
The debug-info component enforces that the `ip` field used to resolve
addresses is a 64bit integer, but the LTTng UST `ip` context field is
32bit on a 32bit cpu. A debug-info component thus fail to recognize that
the trace has all the necessary information to resolve the addresses.

Here is the metadata representing the context on a 32bit system:
        event.context := struct {
                integer { size = 32; align = 8; signed = 1; encoding = none; base = 10; } _vpid;
                integer { size = 32; align = 8; signed = 0; encoding = none; base = 16; } _ip;
        };

Here is the metadata representing the context on a 64bit system:
        event.context := struct {
                integer { size = 32; align = 8; signed = 1; encoding = none; base = 10; } _vpid;
                integer { size = 64; align = 8; signed = 0; encoding = none; base = 16; } _ip;
        };

Solution
========
Remove the integer size check in the
`is_event_common_ctx_dbg_info_compatible()` function.

Drawback
========
None.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I3938a5f71a2406876573a5aa381f7d3dd4d632d6
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1270
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

tests: Run test_bitfield

I noticed that the test_bitfield executable was not run as part of make
check.  Add it to TESTS_LIB.

Change-Id: Id1c40a7916b6326f7655a49e033d5570e40b42e6
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1277
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: duplicate symbol with clang linker

On MacOs with the clang linker, the build fails with:

 duplicate symbol _bt_ctf_value_null in:
     trace-ir/.libs/libtrace-ir.a(trace.o)
     trace-ir/.libs/libtrace-ir.a(trace-class.o)

Make the pointers extern.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I679dc5b647de235a19a958331f7f975c0cb0a08f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1267
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: fd-cache: fd leak on error path

Fixes coverity #1401248

  CID 1401248 (#1 of 1): Resource leak (RESOURCE_LEAK)
  10. leaked_handle: Handle variable fd going out of scope leaks the handle.

Reported-by: Coverity (1401248) Resource leak
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I41662edaa4ff13046e003aec20ac27ff242e338f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1260
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: lib/graph: Dereference before null checks

  CID 1401244 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.source_output_port_added suggests that it may be null,
  but it has already been dereferenced on all paths leading to the check.

  CID 1401260 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.filter_output_port_added suggests that it may be null,
  but it has already been dereferenced on all paths leading to the check.k

  CID 1401243 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.filter_input_port_added suggests that it may be null,
  but it has already been dereferenced on all paths leading to the check.

  CID 1401185 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking graph->listeners.sink_input_port_added
  suggests that it may be null, but it has already been dereferenced on
  all paths leading to the check.

  CID 1401262 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.source_filter_ports_connected suggests that it may be
  null, but it has already been dereferenced on all paths leading to the
  check.

  CID 1401200 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.filter_filter_ports_connected suggests that it may be
  null, but it has already been dereferenced on all paths leading to the
  check.

  CID 1401252 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.source_sink_ports_connected suggests that it may be
  null, but it has already been dereferenced on all paths leading to the
  check.

  CID 1401264 (#1 of 1): Dereference before null check (REVERSE_INULL)
  check_after_deref: Null-checking
  graph->listeners.filter_sink_ports_connected suggests that it may be
  null, but it has already been dereferenced on all paths leading to the
  check.

Reported-by: Coverity (various ids) Dereference before null check
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I88c8172e006e4bae2c7b5e29b501e1a36f19ae5c
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1262
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: cli: value stored is never read

Found with scan-build:
  line 1434, column 4
  Value stored to 'ret' is never read

  line 1281, column 4
  Value stored to 'ret' is never read

  line 1345, column 2
  Value stored to 'ret' is never read

  line 1423, column 3
  Value stored to 'ret' is never read

  line 1396, column 3
  Value stored to 'ret' is never read

Reported-by: scan-build
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I95f06132c34bbf9c4ffbb81423e42cf1834880bd
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1266
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: flt.lttng-utils.debug-info: omitting copy_field_class_content() return value

  ID 1401259 (#1 of 2): Logically dead code (DEADCODE)
  dead_error_line: Execution cannot reach this statement: goto error;.

  CID 1401259 (#2 of 2): Logically dead code (DEADCODE)
  dead_error_line: Execution cannot reach this statement: goto error;.

Reported-by: Coverity (various ids) Logically dead code
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I1b3bb7cba9aec34e6a90386178f99cc505d4da66
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1264
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: flt.lttng-utils.debug-info: memory leak

The gelf_getshdr() function return a pointer to the second parameter on
success and NULL on error. This means that on we overwrite the
`curr_section_hdr` variable and leak memory.

To fix it, we simply change the variable from a heap allocated variable
to a stack allocated variable this has the effect of making code simpler.

Found using scan-build:
  line 334, column 7
  Potential leak of memory pointed to by 'curr_section_hdr'

Report-by: scan-build (ad8e62)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I5cc4049075e4ca1d64d98e17662f935e6562bb6e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1265
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: sink.text.pretty: Unsigned compared against 0

  CID 1401251 (#1 of 1): Unsigned compared against 0 (NO_EFFECT)
  unsigned_compare: This less-than-zero comparison of an unsigned value is
  never true. nr_fields < 0UL.

Reported-by: Coverity (1401251) Unsigned compared against 0
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I8aa8a03a6b03ca1640570d70a51729d001af3b66
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1263
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: src.ctf.fs: init_index_entry() may return uninitialized value

Found with scan-build.
Warning reported: Undefined or garbage value returned to caller

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ifaccdd9f7c8a762fd4eb9f26a5fce4cd092c44df
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1259
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins

Update version to v2.0.0-pre6

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I640770f6def57a88e12beb05a9724c4e3f31cd54

cli/babeltrace.c: update known log level environment variable names

Eventually we could find another way to synchronize the environment
variable names we use throughout the project with the CLI. We want to
decouple the plugins from the CLI (plugins do not know CLI, but CLI
knows plugins) so I don't want the plugins to support an environment
variable defined by the CLI for example.

For the moment, I'm using this:

    rg --pcre2 -N --iglob='*.c' --iglob='*.h' --no-filename -o \
       'BABELTRACE_\w+(?<!CLI)_LOG_LEVEL\b' --color=never | sort | uniq

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I148c5c859570d0b1a654c829b31066155e8ecf5e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1258
Tested-by: jenkins

lib: add unsigned and signed integer value API

This patch replaces the integer value API with the more specific
unsigned and signed integer value API. This opens the whole unsigned
64-bit set for component initialization parameters, query parameters,
queyr results, and, eventually, custom metadata attributes. This is also
more in line with the field API, which has both the unsigned and signed
versions.

Library
=======
The API is simply split in two, where an unsigned integer value contains
a `uint64_t` raw value and a signed integer value is the equivalent of
the previous integer value (`int64_t` raw value).

The types are `BT_VALUE_TYPE_UNSIGNED_INTEGER` and
`BT_VALUE_TYPE_SIGNED_INTEGER`.

There's no interaction between unsigned and signed integer values, in
that, for example, you cannot compare an unsigned integer value with a
signed integer value with bt_value_compare().

Behind the scenes, I kept a single `struct bt_value_integer` for
simplicity. It contains a union of `uint64_t` and `int64_t`. Most
functions call a generic unsigned version, casting to `uint64_t` when
necessary. For example, there's a common bt_value_integer_compare()
which compares the `uint64_t` values.

CLI
===
Before this patch, the `--params` option's format makes any integer
constant a signed integer value object. With this patch, you can create
an unsigned integer value by prepending the constant with `+`. For
example:

    hello=+293, meow=+0x388ab88fd, uint-zero=+0

I wanted to use the `U` suffix, like in C/C++, but the GLib lexical
scanner won't allow a constant to be followed by an identifier without a
whitespace.

The documentation about the format of `--params` is updated for the
CLI's `--help` option and in the man pages.

Plugins
=======
It's up to each plugin to accept, for a given parameter, an unsigned
integer value, a signed integer value, or both. A plugin can be very
strict and it's not a bad thing: there are situations where an unsigned
integer is conceptually required.

The individual changes are:

`src.ctf.fs`:
    In the `trace-info` query result, create unsigned integer values for
    the `id` (stream ID) and `class-id` (stream class ID) entries. A CTF
    metadata ID is never negative.

`src.ctf.lttng-live`:
    In the `sessions` query result, the `timer-us`, `stream-count`, and
    `client-count` entries are still signed integer values. They could
    probably be unsigned, but I want to confirm this with developers
    more involved with this component class before doing the change. For
    the moment, the query operation is unchanged.

`sink.utils.counter`:
    The `step` initialization parameter is now expected to be an
    unsigned integer value. This is never negative.

    I also made the initialization refuse parameters when they don't
    have the expected type. There are new `logging.c` and `logging.h`
    files for this component class because I used logging to communicate
    said errors.

Library tests
=============
The integer value tests now test the signed value API.

I added equivalent unsigned value object tests.

Python bindings
===============
`_IntegerValue` is now the base class of `UnsignedIntegerValue` and
`SignedIntegerValue`. `_IntegerValue` contains the whole public
interface, using template methods with specialized parts defined in
subclasses.

`_IntegerValue` is imported in the `bt2` package for this use case:

    if isinstance(val, bt2._IntegerValue):
        do_something_with(val)

because `int(val)` is always valid.

The bt2.create_value() function still creates a `SignedIntegerValue`
object with an `int` value. This is the safest, as there's no way to
know that this value will be changed to a negative value in the future,
before the value object is frozen. For example, this would not be
possible with a conversion to an `UnsignedIntegerValue`:

    val = bt2.create_value(47)
    val.value = -23

whereas it is possible with a `SignedIntegerValue`. You can still
explicitly create an unsigned integer value:

    counter_comp = my_graph.add_component(counter_cls, 'counter', {
        'step': bt2.UnsignedIntegerValue(29)
    });

Python bindings tests
=====================
`test_value.py` is updated to also test the `UnsignedIntegerValue`
class. _inject_numeric_testing_methods() now accepts an optional
`has_neg` parameter which, if false, does not inject tests which involve
negative integer values.

There is a common `_TestIntegerValue` class which
`SignedIntegerValueTestCase` and `UnsignedIntegerValueTestCase` inherit.
The tests are almost the same, but `SignedIntegerValueTestCase` adds a
few tests related to negative values. Both specific test cases set the
`_CLS` class attribute to the value object class to test, either
`UnsignedIntegerValue` or `SignedIntegerValue`.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ic96ef9e1e16883cb5c59844c6ba5a060936efdb0
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1257
Tested-by: jenkins

lib: value.c: remove BT_VALUE_FROM_CONCRETE() macro

We can just cast to `void *` directly like we do everywhere else in the
library.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: If09f1be2e7f8f9cf3bf964a7d4ada5b3307ef7f4
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1256
CI-Build: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins

lib: move `bt_value` structures to `value-internal.h`

It's unusual to have internal structures in the C file. To remain
consistent with the rest of the library, put them in the internal
header.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I5a946004a172560acecab1b182214293f61f02fd
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1255
CI-Build: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Tested-by: jenkins

bt2: Make test_query_executor pass

This patch fixes and enables test_query_executor.py.

The only important change is in the query method: it is necessary to
check for `is_canceled` prior to calling the API function, since it
asserts that the executor is not cancelled.

I have removed the __eq__ method and tests, since it's quite unlikely to
be useful for the QueryExecutor class.

Change-Id: If88c2552627358b06f7d6068126034a20ebf88e8
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1252
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: Fix Makefile dependency tracking when building out of tree

The problem, from the point of view of the developer:

1. Build out of tree, python bindings enabled.
2. Modify a bindings source file (e.g. trace_class.py,
   native_bt_trace.i).
3. Type "make" in the builddir.
4. The bindings are not rebuilt from the new source files.
5. Sad.

When we build out of tree, it's done in two steps (assume $builddir is
$top_builddir/bindings/python):

1. Copy all the static files (those checked in the repo, not generated
   at runtime) from $srcdir/bt2 to $builddir/bt2.  Those files are placed
   next to other files that are generated (__init__.py, setup.py).
   Together, they form the complete source of the bt2 package.
2. Run setup.py, which runs SWIG and outputs the built package in
   $builddir/build_lib.

The problem described above happens because of how the dependencies are
stated for step #1:

    STATIC_BINDINGS_DEPS = ... bt2/native_bt_trace.i ...
    copy-static-deps.stamp: $(STATIC_BINDINGS_DEPS)

When nothing exists yet in the build directory, bt2/native_bt_trace.i is
resolved to the version in the srcdir (because of make's VPATH, I
believe), so it works fine.  But once the copied files exist, and you
try to change the source files and "make" again, the dependency
bt2/native_bt_trace.i will be resolved to the version in the build dir.
Since that version is not newer than copy-static-deps.stamp, nothing is
done.

The fix it, we can force the dependency to be the one in the source
directory, by prepending $(srcdir) to each file.

Change-Id: Ic09d430e53b59e5afa9ebb19d98ba219dff16537
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1250
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

fix: g_hash_table_insert prior to glib 2.40 returns void

Even the post glib 2.40 version will only return false if the value is
already present in the hashtable but will not fail to add it. The old
value will be handled according to the configuration of the hashtable
which should not be an error.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I5512c921ce220bf8a956c61f594d6a27e6ceb41e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1247
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests/plugins/ctf/Makefile.am: do not set `TESTS` in this file

We want the project's `make check` to continue even if specific tests
fail, so we can't define `TESTS` in a subdirectory's `Makefile.am`,
because if the test fails, `make` fails, and subsequent tests are not
executed.

There are ways to circumvent this, for example using `make`'s
`--always-make` option, but it was decided previously that we want
`make check` to run all tests as is, without any option.

Therefore we set `TESTS` to all tests in `tests/Makefile.am`.

To execute tests specific to a subdirectory, we have the following
targets:

* `check-bindings`
* `check-cli`
* `check-lib`
* `check-plugins`
* `check-python-plugin-provider`

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I1f198ff3ff2b44947042aee7fa5f01759874941f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1254
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: flt.lttng-utils.debug-info: build id note section non-native byte order

Issue
=====
The build id comparison code does not take into account that the target
binary file could be compiled for another architecture and thus could be
of a different byte order. This makes the build_id testcase fail on
PowerPC. The artefacts used for this testcase are compiled on x86 and
are thus in little endian as opposed to PowerPC's big endianness.

Solution
========
Parse the Elf file to extract the endianness of the binary and use that
information to change the byte order of the note section if necessary.

Drawbacks
=========
None.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Id5d761832f9463b38ea0452da36053e885929aa9
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1246
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>

common: add bt_common_read() function

This function wraps the `read()` function to handle EINTR and partial
reads.

It returns the `count` value on success, or a smaller value than `count`
on error.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ib0385855447200d97a58da704dcf68423e857033
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1248
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: common: va_list type is implementation dependant

Issue
=====
BT_ASSERT(*args) happens to work on some library implementations because
the va_list type happens to be a pointer but this is not true with all
libraries. For example, compiling on arm64 results in the following
error:
  In file included from common.c:34:0:
  common.c: In function ‘bt_common_custom_vsnprintf’:
  ../include/babeltrace/assert-internal.h:42:7: error: wrong type
  argument to unary exclamation mark
     if (!(_cond)) {                                                \
         ^
  common.c:1524:2: note: in expansion of macro ‘BT_ASSERT’
    BT_ASSERT(*args);
    ^~~~~~~~~

Solution
========
Remove that BT_ASSERT() call as its intent is unclear anyway.

Drawback
========
None.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Iada5189ef4580e48caefed12835bfab7c23eaf73
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1253
CI-Build: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>
Reviewed-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

Fix: uninitialized variant fc in set_field_refs()

GCC 8.3.0 emits the following warning:

translate-trace-ir-to-ctf-ir.c:935:16: warning: ‘var_fc’ may be used
uninitialized in this function [-Wmaybe-uninitialized]
     named_fc =
     fs_sink_ctf_field_class_variant_borrow_option_by_index(
                     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
           var_fc, i);

I don't think this "condition" is reachable, but it is, in principle
possible. Initializing var_fc to NULL silences this warning.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I1fd8951778e8b3705974324e1cd5d87892c2e418
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1251
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

src.ctf.fs: make trace-info query accept clock-class-offset-{s,ns} parameters

The src.ctf.fs components accept clock-class-offset-s and
clock-class-offset-ns parameters, allowing to add an offset to the
clock.

The trace-info query currently doesn't accept the same parameters.  In
general, I think we'll want the components and trace-info to accept
pretty much the same set of parameters, so that the information reported
by trace-info accurately represents how an eventual component reading
the same traces will work.

This patch splits the parameter handling of the component in a
standalone function re-used by the query.  The advantage of sharing a
function like this is that if we add a parameter to the component, we'll
be forced to think about what happens to the trace-info query.

One behavior change is that trace_info_query now returns
BT_QUERY_STATUS_INVALID_PARAMS if the passed parameters are invalid.

A test is added to test the new functionality.

Change-Id: I3d917b8713fce4ce56a79a6e7c7bdc3e32d9ec0c
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1244
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

ctf: Use g_time_val_from_iso8601 instead of g_date_time_new_from_iso8601

g_date_time_new_from_iso8601 was introduced in glib 2.56, which is quite
recent.  To support build with older glibs, change the code to use
g_time_val_from_iso8601 instead, introduced in glib 2.12.

The code in question only uses this function to validate that the passed
string is ISO8601-compliant, it doesn't use the resulting time value.

The problem of g_time_val_from_iso8601 is that its result is not year
2038-safe on 32-bit systems.  We can expect that by then, that function
will have been removed from glib, forcing us to update our code.

Change-Id: I2f4fdfce9020ecefcfb5d653a3885d343e49bd76
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1245
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests: update EXTRA_DIST of Python bindings

This list is out of date, causing "make dist" to fail.

Sync the list of distributed files with the actual files.

Change-Id: Iedef7411c691689eb6a0505b4ea29d0f319f6860
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1242
Tested-by: jenkins
Reviewed-by: Jonathan Rajotte Julien <jonathan.rajotte-julien@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests: remove non-existent files from EXTRA_DIST

These files were removed in commit

    e8bacbe2bd42 ("Remove stale trace-ir test files")

... so they should be removed from EXTRA_DIST as well.  Currently, it
causes a "make dist" failure:

  make[4]: *** No rule to make target 'test_trace.py', needed by 'distdir'.  Stop.

The tests/lib/trace-ir directory is currently not very useful, as it
contains no tests at all, so we could remove it in theory.  But I
presume we'll add something in there in not too long, so I think we can
leave it there.

Change-Id: Iec44efd7c52e84b35a5a55206055b5d63635fbe1
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1241
Tested-by: jenkins
Reviewed-by: Jonathan Rajotte Julien <jonathan.rajotte-julien@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

include: add missing files to dist tarball

When making a tarball (make dist) and trying to build from it, it fails
because of missing header files.  This patch adds all the necessary
headers to the dist tarball to make "./configure && make" from the
tarball succeed.

Change-Id: I89051cb060e911eac154ef1726034e905b49b085
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1243
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Reviewed-by: Jonathan Rajotte Julien <jonathan.rajotte-julien@efficios.com>

bt2: Add wrapper for bt_plugin_get_version

There is a subtle bug with our use of the "char **OUT" typemap for the
bt_plugin_get_version function.

THe SWIG wrapper does this:

    char *temp_value5;
    arg5 = &temp_value5;

The initial content of temp_value5 is unknown, assume that it points to
unreadable memory.  If the call to bt_plugin_get_version returns
PROPERTY_AVAILABILITY_NOT_AVAILABLE, the value of temp_value5 is
unchanged (but really, it could be anything, again we should assume it
points to unreadable memory).  However, the epilogue of the typemap will
still try to transform the C string to a Python str object.  Doing so
will cause a crash when trying to read the C string.

The fix is to make our own wrapper to bt_plugin_get_version, which
guarantees that if it returns PROPERTY_AVAILABILITY_NOT_AVAILABLE, it
will force the value of the output parameter to be NULL (which is
handled by the epilogue and transformed to None).

Setting the temporary variable to an invalid but non-zero pointer value
((void *) 1) makes this bug consistently reproducible and not rely on
random memory contents.  I suggest we keep it there, in order to quickly
find other instances of this bug in the future.

Change-Id: Idd12ca4c4d4d732b841c1e12420e46e6a5a25874
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1221
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Add git-review config

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: Iece6807efd73611d1a2767c3284fbe6f9f72256c
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1023
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: bt2: enum-conversion warning in native_bt_component_class.i

clang warns about the following enumeration conversion.
  bt2/native_bt_wrap.c:3896:12: warning: implicit conversion from
  enumeration type 'enum bt_message_iterator_status' to different
  enumeration type 'enum bt_self_message_iterator_status'
  [-Wenum-conversion]
                  status = BT_MESSAGE_ITERATOR_STATUS_END;
                         ~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  bt2/native_bt_wrap.c:3898:12: warning: implicit conversion from
  enumeration type 'enum bt_message_iterator_status' to different
  enumeration type 'enum bt_self_message_iterator_status'
  [-Wenum-conversion]
                  status = BT_MESSAGE_ITERATOR_STATUS_AGAIN;
                         ~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  bt2/native_bt_wrap.c:3901:12: warning: implicit conversion from
  enumeration type 'enum bt_message_iterator_status' to different
  enumeration type 'enum bt_self_message_iterator_status'
  [-Wenum-conversion]
                  status = BT_MESSAGE_ITERATOR_STATUS_ERROR;
                         ~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  bt2/native_bt_wrap.c:4513:43: warning: implicit conversion from
  enumeration type 'enum bt_message_iterator_status' to different
  enumeration type 'bt_self_message_iterator_status' (aka 'enum
  bt_self_message_iterator_status') [-Wenum-conversion]
          bt_self_message_iterator_status status = BT_MESSAGE_ITERATOR_STATUS_OK;
                                          ~~~~~~   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  bt2/native_bt_wrap.c:4678:43: warning: implicit conversion from
  enumeration type 'enum bt_message_iterator_status' to different
  enumeration type 'bt_self_message_iterator_status' (aka 'enum
  bt_self_message_iterator_status') [-Wenum-conversion]
          bt_self_message_iterator_status status = BT_MESSAGE_ITERATOR_STATUS_OK;
                                        ~~~~~~   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Icb3bc0141cdd9f35a5a42420bb48167c12e70df0
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1183
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: jenkins

Fix: bt2: int-to-pointer-cast warnings in SWIG utils functions

  bt2/native_bt_wrap.c: In function ‘bt_py3_component_class_query’:
  bt2/native_bt_wrap.c:4453:12: warning: cast to pointer from integer of
    different size [-Wint-to-pointer-cast]
      *result = (void *) PyLong_AsUnsignedLongLong(py_results_addr);
                 ^
  bt2/native_bt_wrap.c: In function ‘bt_py3_component_class_message_iterator_next’:
  bt2/native_bt_wrap.c:4697:3: warning: cast to pointer from integer of
    different size [-Wint-to-pointer-cast]
      (const bt_message *) PyLong_AsUnsignedLongLong(
      ^
  python-plugin-provider.c: In function ‘bt_plugin_from_python_plugin_info’:
  python-plugin-provider.c:341:18: warning: cast to pointer from integer of
    different size [-Wint-to-pointer-cast]
       comp_class = (bt_component_class *)

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I44d15e5d072d993776539eddae5f3e37c6161cac
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1162
CI-Build: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

sink.ctf.fs: set_field_refs(): make `fc` `const` as `fc_type` depends on it

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I9b0a0eb3fc41d9181205cb4f3f33334b90a3b712

Fix: sink.ctf.fs: do not keep different `tgt_fc_type` as `tgt_fc` changes

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ie79306f50018d7f5ca0971e6c9a2410378a4ccc7

lib: bt_graph_{consume,run}(): add unlikely() to unlikely condition

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I37badc4c15fbde16940150854cae1e31305953f1

src.ctf.fs: compute stream range using entire file info group

Given that index entries are sorted within their file info structure and
that file info structures are sorted by time within the file info group,
we can simply compute the range of the stream by using the absolute
first index entry and the absolute last index entry of the entire file
info group.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I5b13451d72a4b1aeb9f31140ebfb025e2d712d7e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1128
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: update bindings to make test_plugins pass

This patch updates the necessary stuff in the bindings to make
test_plugins pass.

One new test is added, test_find_file, which tests passing the path to a
plugin shared object to bt2.find_plugins.

Change-Id: Ie2ac6e830fb18d8839d2879567509cd2f6abbd81
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1087
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: jenkins

bt2: update bindings to make test_component pass

This patch updates component.py such that the test_component test
passes.

With the API, it is no longer possible to obtain the graph from a
component, so things related to this are removed from both files.

Change-Id: I3ab0e6aaaa8ec1ad7378abb78072e6e130e1c2a4
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1085
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: update bindings to make test_component_class pass

This patch updates the binding with the goal of making
test_component_class pass.

Besides the obviously needed changes to the component class code, there
are a bunch of changes to things around it (component, graph), just
enough to make the test run.

One notable change is the _SharedObject._release method, used to make the
Python object stop manage the reference it currently owns.

Another functional change is how the component classes _query method
reports that it's not implemented.  Currently, it default to return the
NotImplemented object.  Instead, I made it raise a NotImplementedError.
Returning NotImplemented is useful if the caller has an alternative to
try when the method is not implemented.  In our case, there is nothing
to do if the specific class doesn't implement _query.  The exception is
caught in bt_py3_component_class_query, where other types of exceptions
need to be handled anyway.

Some changes are required to the library, to be able to get a specific
component class type from a specific component (e.g. a
bt_component_class_source from a bt_component_source).

Change-Id: I895817bc996c21b6a9c6d57605cc64f9ff61dfcc
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1083
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: Add functions to borrow specialized component classes from specialized components

The API currently contains a function to obtain a 'bt_component_class *'
from a 'bt_component *'.  This patch adds functions to obtain a
specialized component class from a specialized component.  For example,
a 'bt_component_class_source *' from a 'bt_component_source *'.

Change-Id: Ide192c7759f630adbc884def2e5c462ce2b8f973
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1161
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: update value.py, make test_value pass

This patch updates value.py to match the current API.

The only known thing missing (to be done later) is to split classes in
their const and non-const variants (e.g. BoolValue and BoolValueConst),
where the methods that can modify values are only present on the
non-const versions.

The test is largely unchanged, except for the removal of things related
to frozen values.

Change-Id: I6424b0e6208ffe2cbe5e60fdd08579326abff29a
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1022
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: update object model

This patch updates the Babeltrace Python object model (what all
Babeltrace Python objects inherit from) to match the current status of
the API.

We have both Shared and Unique objects.  Both types have a _ptr
member, a SWIG Python object wrapping a typed pointer (e.g. pointer to
struct bt_value).

A Shared object type is defined by inheriting from _SharedObject.  It
wraps and owns a reference to a Babeltrace object that is reference
counted.  The reference count is dropped when the Python object gets
deleted.  The _SharedObject base class defines two class methods,
_get_ref and _put_ref, which must be implemented by subclasses.  These
implementations must respectively get and put a reference of the
specialized pointer (_ptr) they wrap.

A Unique object type is defined by inheriting from _UniqueObject.  It is
not refcounted.  Rather, it is wholly owned by another object (the
owner) that is itself refcounted.  A Unique object, in the Babeltrace
API, is destroyed when its parent is destroyed.  There is therefore a
risk that someone would keep a (Python) reference to a unique object
that got destroyed, because its parent got destroyed, resulting in some
use-after-free bug.  To avoid this, we make it so that when a Python
object representing a Babeltrace Unique object is created, that Python
object obtains a reference on the owner Shared object.  When the Unique
Python object is destroyed, the reference on the Shared parent is
dropped.

The concepts of Freezable and Private objects don't exist anymore, so
they are removed.  I have also removed the _PrivateConnection class,
since it won't be needed in the end and inherited from _PrivateObject.
For other classes representing Babeltrace concepts, I have made them
inherit the object type they'll need to inherit in the end (when all the
Python bindings will be fixed), even though none of these classes is
functional at the moment.

Change-Id: I9ab743dd7a28407913945648c1d74c467550df14
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1021
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: string format warnings on 32 bits

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I49fdb170642bfd0c7326719f42973863af41c8e7
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1141
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: unused-function warnings in lib/graph/iterator.c

clang warnings:
  iterator.c:87:6: error: unused function '_set_self_comp_port_input_msg_iterator_state' [-Werror,-Wunused-function]
  void _set_self_comp_port_input_msg_iterator_state(
       ^
  iterator.c:885:6: error: unused function '_set_iterator_state_after_seeking' [-Werror,-Wunused-function]
  void _set_iterator_state_after_seeking(
       ^

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia409f1dd696cdfed58de23aad926c4effdc26307
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1034
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: may-uninitialized warning in trace-ir-metadata-field-class-copy.c

GCC warning:
  In file included from ../../../include/babeltrace/common-internal.h:30:0,
                   from trace-ir-metadata-field-class-copy.c:31:
  trace-ir-metadata-field-class-copy.c: In function ‘copy_field_class_content_internal’:
  ../../../include/babeltrace/trace-ir/field-class-const.h:196:3: error:
  ‘out_tag_field_class’ may be used uninitialized in this function
  [-Werror=maybe-uninitialized]
     bt_field_class_put_ref(_var);  \
     ^~~~~~~~~~~~~~~~~~~~~~
  trace-ir-metadata-field-class-copy.c:421:18: note: ‘out_tag_field_class’ was declared here
    bt_field_class *out_tag_field_class;

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I73c6057f9795b153f95c1287f3f3f8e529328fbf
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1033
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: maybe-uninitialized warnings in translate-trace-ir-to-ctf-ir.c

GCC warning example:
  translate-trace-ir-to-ctf-ir.c:929:14: error: ‘struct_fc’ may be used
  uninitialized in this function [-Werror=maybe-uninitialized]
       named_fc = fs_sink_ctf_field_class_struct_borrow_member_by_index(
       ~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
       struct_fc, i);
       ~~~~~~~~~~~~~

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I24ba8095f26cec640e0a24473b4fee05d1c404af
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1032
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: maybe-uninitialized warning in create_relative_field_ref()

  translate-trace-ir-to-ctf-ir.c: In function ‘resolve_field_class’:
  translate-trace-ir-to-ctf-ir.c:302:6: error: ‘tgt_fc_name’ may be used
    uninitialized in this function [-Werror=maybe-uninitialized]
       g_string_assign(tgt_field_ref,
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
         tgt_fc_name);
         ~~~~~~~~~~~~
  translate-trace-ir-to-ctf-ir.c:177:14: note: ‘tgt_fc_name’ was declared here
    const char *tgt_fc_name;

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I8d3398a852bf3ca412bb479eae72a81a3959d220
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1031
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: unused-variable warnings in lib/graph/iterator.c

GCC warnings:
  make[3]: Entering directory '/home/frdeso/projets/babeltrace/optimalize/lib/graph'
    CC       iterator.lo
  iterator.c: In function ‘_set_iterator_state_after_seeking’:
  iterator.c:889:59: error: variable ‘new_state’ set but not used [-Werror=unused-but-set-variable]
    enum bt_self_component_port_input_message_iterator_state new_state = 0;
                                                             ^~~~~~~~~
  iterator.c: In function ‘find_message_ge_ns_from_origin’:
  iterator.c:1159:59: error: unused variable ‘init_state’ [-Werror=unused-variable]
    enum bt_self_component_port_input_message_iterator_state init_state =

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia14a89f330198e4109af14f755aeed3378912256
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1030
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: maybe-uninitialized warning in bt_ctf_stream_flush()

GCC warning:
  In file included from ../../include/babeltrace/ctf-writer/fields-internal.h:41:0,
                   from ../../include/babeltrace/ctf-writer/field-wrapper-internal.h:26,
                   from ../../include/babeltrace/ctf-writer/event-internal.h:37,
                   from stream.c:32:
  stream.c: In function ‘bt_ctf_stream_flush’:
  ../../include/babeltrace/ctfser-internal.h:568:36: error:
  ‘packet_context_offset_bits’ may be used uninitialized in this function
  [-Werror=maybe-uninitialized]
    ctfser->offset_in_cur_packet_bits = offset_bits;
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~
  stream.c:1608:11: note: ‘packet_context_offset_bits’ was declared here
    uint64_t packet_context_offset_bits;

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia7457b1a444290ba1a71459bfaf707812227899c
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1029
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: unused-variable warnings in bt_X_freeze() when not in DEV_MODE

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Idf9e91b2abd9b2720343917c2d3ffdf207f82735
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1028
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: src.ctf.fs: metadata-info: sanitize `path` param

Issue
=====
Omitting to pass the mandatory `path` parameter to the `metadata-info`
query results in a BT_ASSERT() failure when calling
bt_value_get_string(), or a NULL pointer dereference in non-DEV_MODE.

Similarly, setting the `path` to a non-string value would also result in
a BT_ASSERT() failure, or a invalid memory access in non-DEV_MODE

Solution
========
Confirm that the `path` parameter is present and is of the string type and
print error messages accordingly it's not.

Known drawbacks
===============
None.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I8182de2797c375262d77a4a8961bedb4c5ef9578
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1081
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: flt.lttng-utils.debug-info: extra colon in `bin` member

There is an extra ':' character in the `bin` member of the `debug-info`
field.

Example of the issue from the tests:
   debug_info = { bin = "libhello_so:+0x15a6", func = "bar+0xa9", src =
   "libhello.c:13" }

The "libhello_so:+0x15a6" will now be printed "libhello_so+0x15a6".

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ie7c4ea54e534eeb8a1bdd503760b791a01aa9594
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1129
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

debug info: Skip test that depends on Python bindings

This test depends on the Python bindings, which don't work right now.
Skip it, so that it doesn't cause a "make check" failure while we are
repairing the bindings.

Change-Id: I8e9b1092e2794875b249b060581f4305be640a88
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1126
Tested-by: jenkins
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

babeltrace: Remove leftovers

It seems like these files should have been removed when removing the
babeltrace Python bindings in commit

    1c81bd8895c0 ("Remove babeltrace 1 Python bindings and tests")

Right now, the test causes a "make check" failure, because it tries to
import the babeltrace Python package, which no longer exists.

Change-Id: Ied4871ec23444776fca601b6f78b583ef8137c81
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1125
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

doc: Make python bindings doc build

Babeltrace 1 Python bindings (the babeltrace package) were removed in
commit

    1c81bd8895c0 ("Remove babeltrace 1 Python bindings and tests")

However, the Python doc still references it.  It tries to import it to
get the version, and therefore fails to build.  Update it to import the
bt2 package instead.

Change-Id: I656a89ad83069f6794abf4c775a14fb9be2ecba0
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1124
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

python-plugin-provider: Make it build

This patch fixes building with --enable-python-plugins.

One problem is that the python plugin provider uses some
functions coming from the internal headers of the lib, which use some
logging functions that are internal to the lib, but not included by
those internal headers.  Trying to add the relevant include file
(lib-logging-internal.h) conflicts with python-plugin-provider's own
logger.

python-plugin-provider doesn't really need to have its own logger.  Even
though it is built as a separate shared object (so that we can package
it separately), it can be considered as internal to the library (and
that's why it includes some internal lib headers).  Therefore, the
easiest way to fix this is to get rid of its logger and make the lib
header files that use lib logging include the required header file
(lib-logging-internal.h).

This requires the python-plugin-provider shared object to access a few
additional symbols not currently exposed by libbabeltrace, which
requires the removal of a few BT_HIDDEN.

I have also included fixes to the python-plugin-provider which I have
on hand.  However, it won't work until the Python bindings are up and
running, so I have marked it as skipped.

Change-Id: I9b46d91fc635d670fef2e4d3c8b32203ec34443f
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1101
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Remove clock_class_priority_map.py

This concept no longer exists in the BT api, and therefore this file
won't be needed anymore.

Change-Id: Ib298f5463ee701be3a0a5b3d4b37b6e7ecf8d5b3
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452469
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: headers: make static inline upcasts compatible with C++

Issue
=====
There are many static inline functions which upcast a pointer from its
original type to another one known to be a parent. The functions upcast
with `(void *)` or `(const void *)`. While this works in C, it does not
in C++, which is more strict:

    In file included from install/include/babeltrace/babeltrace.h:75,
                     from test.cpp:1:

    install/include/babeltrace/trace-ir/field-class-const.h: In function
    ‘const bt_field_class_enumeration_mapping*
    bt_field_class_unsigned_enumeration_mapping_as_mapping_const(const
    bt_field_class_unsigned_enumeration_mapping*)’:

    install/include/babeltrace/trace-ir/field-class-const.h:100:9:
    error: invalid conversion from ‘const void*’ to ‘const
    bt_field_class_enumeration_mapping*’ [-fpermissive]

      return (const void *) mapping;

Solution
========
Use the new __BT_UPCAST() and __BT_UPCAST_CONST() macros to upcast
within headers. Those macros are defined in `babeltrace/types.h`. We use
static_cast<>() in C++ with the specific destination type, going through
`void *` or `const void *` to keep the same address.

Known drawbacks
===============
None.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Iaac8e55635eb2f358a546c32bd886a484f5ebfeb
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452302
Reviewed-by: Simon Marchi <simon.marchi@polymtl.ca>

Fix: headers: add missing end of `extern "C"` curly brackets (for C++)

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I5f6277ce3075f26bec6507b7b26977d33aea858f
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452301
Reviewed-by: Simon Marchi <simon.marchi@polymtl.ca>

Fix: include/Makefile.am: add missing `babeltrace/ctf-writer/object.h`

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I29700d6c62b0ebd910b6a6715ab7edbc2126168f

bt2: Rename files to use singular form

I am mildly annoyed that most Python files use the singular form of the
concept they are related too (component.py, event.py) but others use the
plural form (fields.py, values.py).

Change this to consistently use the singular.

Change-Id: I1a33c47ec45ef3359025e0a66f01eaefc09fecdf
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452257
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: Mass clock_value -> clock_snapshot rename

The clock_value concept has been renamed to clock_snapshot.  Do a mass
rename in the bt2 Python bindings and the associated tests.

Change-Id: I2b275f554c20a12a0f1891c4deed0046e4d3e8e6
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452254
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: Mass field_types -> field_class rename

The concept of field_type has been renamed to field_class.  This patch
does a mostly mechanical rename across the Python bindings and their
tests.

Change-Id: I3c2a870262e3cfeddaec7d3565eb635a4c70d131
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452253
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: Mass notification -> message rename

As the "notification" concept has been renamed to "message", I decided
to do a mass rename (including renaming files).  It's quite
straightforward and will reduce a little bit the noise in further
patches.

The changes were done mostly mechanically, but I verified them to make
sure it didn't go too wild (like changing "notifier" to "msgier", which
we wouldn't want).

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I6e6dec1234864400a2570a4726c32a1398cf42c0
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452232
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: Sync native_bt_field_class.i with field-class-const.h

Commit

    2d97b62eea05 ("lib: have dedicated, unique unsigned and signed enum FC mapping objects")

Changed some declarations in field-class-const.h without doing the
corresponding changes in native_bt_field_class.i.  I am now seeing this
build error:

    bt2/native_bt_wrap.c: In function ‘_wrap_field_class_unsigned_enumeration_borrow_mapping_by_index_const’:
    bt2/native_bt_wrap.c:11087:3: error: unknown type name ‘bt_field_class_unsigned_enumeration_mapping_ranges’; did you mean ‘bt_field_class_unsigned_enumeration_mapping’?
       bt_field_class_unsigned_enumeration_mapping_ranges **arg4 = (bt_field_class_unsigned_enumeration_mapping_ranges **) 0 ;
       ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
       bt_field_class_unsigned_enumeration_mapping

Change-Id: I897749a6a7fe5bd1c3c16b1ea9625820e77f112b
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452225
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: have dedicated, unique unsigned and signed enum FC mapping objects

This patch makes the enumeration field class mapping API aligned with
the structure field class member and variant field class option APIs. In
other words, a signed or an unsigned enumeration field class mapping is
now a dedicated type. An enumeration field class mapping object is
unique and always belongs to its parent field class.

bt_field_class_unsigned_enumeration_borrow_mapping_by_index_const() and
bt_field_class_signed_enumeration_borrow_mapping_by_index_const() now
return borrowed enumeration field class mapping objects.

You can use
bt_field_class_unsigned_enumeration_mapping_as_mapping_const() to change
a signed or unsigned enumeration field class mapping into a generic
enumeration field class mapping. The API for a generic enumeration field
class mapping is:

    const char *bt_field_class_enumeration_mapping_get_label(
        const bt_field_class_enumeration_mapping *mapping);

    uint64_t bt_field_class_enumeration_mapping_get_range_count(
        const bt_field_class_enumeration_mapping *mapping);

You cannot create an enumeration field class mapping object: you still
map a name to a specific range with
bt_field_class_unsigned_enumeration_map_range() and
bt_field_class_signed_enumeration_map_range().

This patch also makes the `flt.lttng-utils.debug-info` and `sink.ctf.fs`
component classes use the updated API.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Iba5da648ca69ec0ca0c5df09a9f383dd4b1e3972
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/451996

tests/bindings/python: Mark all tests as skipped

This patch makes us skip all the Python bindings tests.  The goal is for
these commands to run with success (even though they doesn't actually test
anything):

    $ tests/bindings/python/bt2/test_python_bt2
    $ tests/bindings/python/babeltrace/test_python_babeltrace

The strategy will be to gradually fix the bindings and their tests, and
gradually remove those skip annotations.  At any point, we should be
able to run the testsuite with success, ensuring that we don't regress.

After this patch, "make check" is clean for me, with python bindings
enabled (--enable-python-bindings).

I am not sure what is still relevant in the python/babeltrace
directory, I suppose the ctf writer stuff still is.  I think we can sort
this out later.

Change-Id: I1169c255c0c7c29fa5de75a358c02c4cd57aae20
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452155
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

bindings/python/bt2: Make the bt2 package importable

This patch changes the minimum necessary to make the bt2 package
importable.  In other words, for this command to succeed:

    $ tests/utils/test_python_bt2_env python3 -c "import bt2"

In some cases, I've replaced some references to concepts that don't
exist with the new concepts (e.g. some NOTIFICATION_TYPE_* to
MESSAGE_TYPE_*).  In other cases, (e.g. fields.py and field_types.py),
it was easier to remove the code, knowing we'll add it back when
tackling the tests that require these files.

After this milestone, we'll be able to at least run some tests that
import the bt2 package, so we'll be able to make sure we don't regress
things while we fix others.

Change-Id: I930d5a4dd129777cb40c607670e27002bb727e39
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452154
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: Fix function name to match declaration

Commit

  c4cae502570e ("lib: Fix function name to match declaration")

used an erroneous name for a function definition, this patches fixes it.

Change-Id: I4190b60c9f6db62bf6b50fd095c402494d596a27
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.gerrithub.io/c/eepp/babeltrace/+/452153
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: Philippe Proulx <eeppeliteloop@gmail.com>