drtracing.org Git - babeltrace.git/log

Fix: format string warnings on mingw

Since we have added format string attributes in the tap.h file, we get
these diagnostics on Windows:

      CC       test_bitfield.o
    test_bitfield.c: In function 'run_test_unsigned_write':
    test_bitfield.c:51:36: error: unknown conversion type character 'l' in format [-Werror=format=]
       51 | #define DIAG_FMT_STR(val_type_fmt) "Failed reading value written \"%s\"-wise, with start=%i" \
          |                                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    test_bitfield.c:51:36: note: in definition of macro 'DIAG_FMT_STR'
       51 | #define DIAG_FMT_STR(val_type_fmt) "Failed reading value written \"%s\"-wise, with start=%i" \
          |                                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    test_bitfield.c:173:8: note: in expansion of macro 'check_result'
      173 |    if (check_result(src_ui, readval, target.c, unsigned char,
          |        ^~~~~~~~~~~~
    test_bitfield.c:175:9: note: format string is defined here
      175 |      "%llX")) {
          |         ^

By default, printf on Windows doesn't know about the `ll` format string
length modifier (as in %llx).  However, we build Babeltrace with the
__USE_MINGW_ANSI_STDIO macro defined, which makes mingw use a
replacement printf implementation that knows about `ll`.

When we define our own functions with a format attribute, we must then
use __MINGW_PRINTF_FORMAT instead of `printf`.  __MINGW_PRINTF_FORMAT
will take the right value, depending on __USE_MINGW_ANSI_STDIO, so it
will validate the format string according to the selected printf
implementation.

Change-Id: I89b194e915250a0e0a617817a1cd10fedb156639
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2161
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

Fix: lib: field-class.c: memory leak on error

scan-build report:
Potential leak of memory pointed to by 'var_with_sel_fc'
File: src/lib/trace-ir/field-class.c Line: 1524

Reported-by: scan-build
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I6f52562c8549810ffb82c6ae4b34707ad6a32576
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2158
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Cleanup: lib: field-class.c: fix logging message for variant FC

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I5b23f843493d7c68121c262e5cbd795503996ad9
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2159
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Require automake >= 1.12

The test suite LOG_DRIVER statement requires that automake >= 1.12 be used
during bootstrap.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I848324bfabcf158f53bed5ed4c1045546e9a8ce5
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2152
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: fix wint_t warning on MSYS2/mingw

We get the following warning when compiling on Windows:

    common.c: In function 'handle_conversion_specifier_std':
    common.c:1419:30: warning: 'wint_t' {aka 'short unsigned int'} is promoted to 'int' when passed through '...'
     1419 |    BUF_STD_APPEND_SINGLE_ARG(wint_t);
          |                              ^
    common.c:1419:4: note: in expansion of macro 'BUF_STD_APPEND_SINGLE_ARG'
     1419 |    BUF_STD_APPEND_SINGLE_ARG(wint_t);
          |    ^~~~~~~~~~~~~~~~~~~~~~~~~
    common.c:1419:30: note: (so you should pass 'int' not 'wint_t' {aka 'short unsigned int'} to 'va_arg')
     1419 |    BUF_STD_APPEND_SINGLE_ARG(wint_t);
          |                              ^

When expanded, this problematic macro usage contains:

    wint_t _arg = va_arg(*args, wint_t);

I think the suggestion of passing `int`, not `wint_t` makes sense.  This
is because any integral argument smaller than an int would have already
been promoted to int when passed to the variable arguments function
that eventually called handle_conversion_specifier_std.  So it makes
sense to fetch this variable as an int.

We don't see this warning on x86-64 Linux because there, sizeof(wint_t)
is 4, the same as an int.

Change-Id: Ia1519b15a56a58022254be98597c1f3e020db9b9
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2160
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

src.ctf.lttng-live: make lttng_live_component_create static

This function is not used outside this file, so make it static.

Change-Id: If2836f7eea7c8243d26088b4e026b31a4ed8c82d
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2139
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Cleanup: flt.lttng-utils.debug-info: coding style

When I first wrote this component class, I was using the wrong coding
style standard. More specifically, the main problem is that the number
of tabs on breaking lines is wrong at bunch places in the debug-info
files.

I am currently working on this code and it annoyed me. So I fixed most
of it.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I41b3d6b3a35546e5ce4f65470bed1303966523c4
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2134
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

port: fix link order of babeltrace2.exe

On both Cygwin and MINGW, the link order of static libraries in an exe
requires the dependent libraries to come first. In this case
'libbabeltrace2-param-parse.la' depends on 'libbabeltrace2.la' for the
'bt_value_bool_create_init' symbol and thus has to be listed before.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Change-Id: I64df3dddcc65233c63f61af3fe7a6a093990c203
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2153
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

Fix: lib: field.c: calling _PUT_REF() on unique objects on error

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I91b22cc4afa492f56cf2a1f460bb29dd61c96c08
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2150
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: rename `bt_*_compare()` to `bt_*_is_equal()`

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I7782ad5c7b5374e88d5f0e31c4fe3fec4e4bea6c
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2145
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Cleanup: lib: graph.c: dead assignment

Reported-by: Scan-build - Dead assignment
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I512551f8ca951eeb004c0a819e7123c363eb5036
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2149
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: fix some warnings shown with -Wextra on gcc 9.1.0

Fix this:

    /home/simark/src/babeltrace/src/ctf-writer/event.c: In function ‘bt_ctf_event_create’:
    /home/simark/src/babeltrace/src/ctf-writer/event.c:626:3: error: cast between incompatible function types from ‘void (*)(struct bt_ctf_field_wrapper *)’ to ‘void (*)(void *, void *)’ [-Werror=cast-function-type]
      626 |   (release_header_field_func) destroy_event_header_field);
          |   ^

by adding a second parameter, stream_class, to destroy_event_header_field, to
match the release_header_field_func type (which takes two pointers).

Fix this:

    /home/simark/src/babeltrace/src/lib/trace-ir/clock-class.c: In function ‘bt_clock_class_create’:
    /home/simark/src/babeltrace/src/lib/trace-ir/clock-class.c:125:3: error: cast between incompatible function types from ‘void (*)(struct bt_clock_snapshot *, struct bt_clock_class *)’ to ‘void * (*)(void *, void *)’ [-Werror=cast-function-type]
      125 |   (bt_object_pool_destroy_object_func)
          |   ^

by making bt_object_pool_destroy_object_func return `void` instead of `void *`
(none of the implementations actually return anything, it seems like a copy
paste error from the typedef above).

Change-Id: I8707a4a3b3eb34d84604f9b8117a20eaa5f50c83
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2138
Tested-by: jenkins <jenkins@lttng.org>

tests: test_output_path_ctf_non_lttng_trace_output small improvements

- Give a name to temporary directories
- Use bt_cli, so the executed command lines are printed

Change-Id: Ife9cadbabeca86a58ac976dc61b7c4c8d25c6efb
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2143
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Move CLI params parsing to its own lib

This will allow using it for parameter parsing in a test, for
convenience.

The sharp-eyed reader will have noticed that this patch also removes
unused structures from src/cli/babeltrace2-cfg-cli-args.c. These were
likely forgotten there when the parameter parsing code was moved to its
own file.

Change-Id: Ibc7ecb238905051492718f1d176c0fb5f0c172a8
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2135
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Cleanup: flt.lttng-utils.debug-info: remove leftover comments

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I98412f2f27d8b3c7611f26021b1e8dca12062b87
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2131
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: tests: remove unused imports

Change-Id: I5969e33915073d9638dbdcc0ec6c49afec81d5e7
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2144

cli: update error message in set_stream_intersections

The error message printed when the `babeltrace.trace-infos` query fails,
regardless of the reason, is "Component class does not support the
`babeltrace.trace-infos` query".  I think it's misleading, because this
is just one particular error case.  It could be that the component class
supports the query, but failed because it didn't like the parameters we
passed.

Update the message to be a bit more generic.  The fail_reason printed
just after gives a bit more precision about the error.

Change-Id: If556129b612cb6aa9c480b3f863368e48e8a6d4a
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2115
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests: get rid of missing-field-initializers warning in test_bin_info.c

This is not super important, and the current code looks correct, but
clang produces this warning:

    /home/smarchi/src/babeltrace/tests/plugins/flt.lttng-utils.debug-info/test_bin_info.c:93:7:
    error: missing field 'short_name' initializer
    [-Werror,-Wmissing-field-initializers]
            {NULL}};

I'd like to have this warning enabled, since it can catch some
legitimate mistakes, so I thought it would be useful to change the code
to avoid having a warning.

Change-Id: Icb5d3227b87ba1bbf405cd320f13886bf313492f
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2133
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests: constify format strings in tap.h

Format strings are typically read-only data (literal strings), and they
are not modified by the functions receiving them, so it makes no sense
for them not to be const.

This helps avoid some "passing value to function foo discards const
qualifier" kind of warnings.

Change-Id: Ibaaaf5690bed21e1d541f1b676627dfd70541219
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2130
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: jenkins <jenkins@lttng.org>

tests: add format attributes to functions receiving format strings in tap.h

This makes the compiler validate the format strings against the types of
the passed arguments. It found a few (minor) issues, which are also
fixed by this patch.

Note that in test_bitfield.c, the format string now includes "0x" before
printing the number in hexadecimal, to avoid any confusion when the user
reads it.

Change-Id: I07cac88aa3cdd445d79f2c12bc0f9333f6a768a9
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2129
Tested-by: jenkins <jenkins@lttng.org>

trace{-class}.c: BT_ASSERT_PRE() -> BT_ASSERT_POST() (ref. count check)

Those are postcondition assertions because they check some state after a
user function returns.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I3705ee451ff8ae024a2eb17fbe6ae9cc2b11d05e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2127
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib: rename `bt_object_{get,put}_no` -> `bt_object_{get,put}_ref_no`

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I5747bdb748d585b25e0ffbe4701b9aef5434728e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2126
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

test_package.py: use list of public names instead of dedicated methods

This patch changes `test_package.py` so that the public names to verify
are listed in the `_public_names` list. The code creates the
corresponding test methods.

This makes it less cumbersome to add public name tests in the future.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I0d370d4b5fd4564000e1e0f273d3b1b36a8acc43
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2125
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

field_class.py: normalize field class names (`_NAME`)

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ib51e612dc3c6b2610bf9dc0c987ce4816e3bdd9f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2124
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib, bt2: rename "signed/unsigned sel." -> "integer signed/unsigned sel."

This matches the option field class terminology and makes it possible to
add other types of variant selectors in the future.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I8b1d5b10ffb9062b09ac5ce2ef6607045cdfb5cb
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2123
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib: add option field classes with integer selectors

This patch adds two new option field classes which have unsigned and
signed integer field classes as selectors. The purpose of this patch is
to be able to use the same selector for an option field class and a
contained, optional variant field class so that you can save space. For
example (pseudo-TSDL):

    enum : integer { size = 8; } {
        NONE,
        BANANA,
        APPLE,
    } tag;

    option <tag> {
        variant <tag> {
            string BANANA <1>;
            int APPLE <2>;
        } <1...0xff>;
    } opt_var;

Above, the optional variant field does not exist when the `tag` field is
0. If `tag` is anything else, then the variant field exists, and the
selected variant field's option also depends on this same value (1 and 2
select existing options).

Library changes
===============
There are now four option field classes:

Option without selector:
    Create with:

        bt_field_class *bt_field_class_option_without_selector_create(
            bt_trace_class *trace_class,
            bt_field_class *content_field_class);

Option with boolean selector:
    Create with:

        bt_field_class *bt_field_class_option_with_selector_bool_create(
            bt_trace_class *trace_class,
            bt_field_class *content_field_class,
            bt_field_class *selector_field_class);

    `selector_field_class` must be a boolean field class.

    By default, the selector is not reversed, in that when the selector
    boolean field is true, the option's field exists.

    You can reverse this logic with:

        void bt_field_class_option_with_selector_bool_set_selector_is_reversed(
            bt_field_class *field_class, bt_bool selector_is_reversed);

    You can get this property with:

        bt_bool
        bt_field_class_option_with_selector_bool_selector_is_reversed(
            const bt_field_class *field_class);

Option with unsigned integer selector:
Option with signed integer selector:
    Create with one of:

        bt_field_class *
        bt_field_class_option_with_selector_integer_unsigned_create(
            bt_trace_class *trace_class,
            bt_field_class *content_field_class,
            bt_field_class *selector_field_class,
            const bt_integer_range_set_unsigned *range_set);

        bt_field_class *
        bt_field_class_option_with_selector_integer_signed_create(
            bt_trace_class *trace_class,
            bt_field_class *content_field_class,
            bt_field_class *selector_field_class,
            const bt_integer_range_set_signed *range_set);

    For both versions:

    * `selector_field_class` must be an integer field class.

    * `range_set` tells, for such an option field, which values of the
      selector field make it contain the optional field.

    * `range_set` must contain at least one integer range.

    * On success, the function freezes `range_set`.

    You can borrow the integer range sets with:

        const bt_integer_range_set_unsigned *
        bt_field_class_option_with_selector_integer_unsigned_borrow_ranges_const(
            const bt_field_class *field_class);

        const bt_integer_range_set_signed *
        bt_field_class_option_with_selector_integer_signed_borrow_ranges_const(
            const bt_field_class *field_class);

Python bindings changes
=======================
There are new Python classes in `field_class.py` to wrap the new field
classes.

To create the new field classes, use:

* _TraceClass.create_option_without_selector_field_class()
* _TraceClass.create_option_with_bool_selector_field_class()
* _TraceClass.create_option_with_integer_selector_field_class()

_TraceClass.create_option_with_integer_selector_field_class() relies on
the type of `selector_fc` to create either an option with unsigned
integer selector field class or an option with signed integer selector
field class.

Notable component class changes
===============================
`flt.lttng-utils.debug-info`:
    Copies all option field classes and their properties.

`sink.text.details`:
    Writes all the option field class properties:

    * Whether or not the selector is reversed for an option with boolean
      selector field class.

      It looks like this:

          opt: Option (boolean selector) (Selector field path [Event payload: 2]):
            Selector is reversed: Yes
            Content: Unsigned integer (64-bit, Base 10)

    * The sorted integer ranges for an option with integer selector
      field class.

      It looks like this:

          opt: Option (signed integer selector) (Selector field path [Event payload: 1]):
            Selector ranges: [1, 1548] [1634] [2960, 3505]
            Content: Unsigned integer (64-bit, Base 10)

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I31e1deba974bf30274d567a73a00ea80d028f7ae
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2122
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

fs-sink-trace.c: lttng_validate_datetime(): ignore deprecated decl.

Disable `deprecated-declarations` compiler warnings for
lttng_validate_datetime() because we're using `GTimeVal` and
g_time_val_from_iso8601() which are deprecated since GLib 2.56
(Babeltrace supports older versions too).

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I4da3bda550b3fd7d7ede13ff3b78ec78ad6f7839
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2121
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Fix: iterator.c: initialize `status` (may be used uninitialized)

From this GCC warning:

    iterator.c: In function ‘find_message_ge_ns_from_origin’:
    iterator.c:1611:9: error: ‘status’ may be used uninitialized in this
                              function [-Werror=maybe-uninitialized]
     1611 |  return status;
          |

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: If6f5f8a63d5a530284a08b9c6dd7f67ecfc9a818
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2120
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

m4: sync ax_pthread.m4 with autoconf archive

Sync with [1].

In particular, this lets us configure with -Wunused-but-set-parameter.

[1] https://raw.githubusercontent.com/autoconf-archive/autoconf-archive/62e8491dbc174113a04fe910cad1e92e8a9e2164/m4/ax_pthread.m4

Change-Id: I496e3d9a8e896b04bf3cc87ddd41b1b144d23ee5
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2128
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

bt2: add `IntegerRangeSet` const classes

Outside of simple rename and splitting, this commit changes the `__eq__`
methods so that it's possible to compare const and non-const objects
together.

I also renamed the `_range_type` to `_range_pycls` to follow what is
done in other files and `_borrow_range_by_index_ptr` to
`_borrow_range_ptr_by_index` to increase readability.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I04b4027d297d088c6531be04681598653601579d
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2118
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Tested-by: jenkins <jenkins@lttng.org>

Tests: bt2: move `_create_const_field()` to utils.py

So it can be used by other test cases. It will be used by the
`test_integer_range_set.py` file in a following commit.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ic64a60e9d1a7bca11292c824531949602d567a6f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2117
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Cleanup: bt2: typo "nsigned" -> "unsigned"

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Id349adc7d4b1dfedadad2a1c014141b5e408b9bb
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2116
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: make it mandatory to have seek_X if can_seek_X is defined

It's currently possible for a component class to provide can_seek_X
(can_seek_beginning or can_seek_ns_from_origin) without the
corresponding seek_X.  This doesn't make much sense, as if can_seek_X
returns true, Babeltrace assumes that seek_X can be called (which is
inconvenient if it's not provided).  That only leaves room for the case
where can_seek_X always returns false, which is not useful.

This patch makes it only possible to provide a can_seek_X if the
corresponding seek_X is provided, for both C and Python user component
classes.  It's still possible, however, to provide seek_X without
can_seek_X, in which case Babeltrace assumes that it's always possible
to call seek_X.

In the C API, component class method setters for seek_X and can_seek_X
are merged in a single function that sets both.  Since we assert that
seek_X is not NULL, this ensures that can_seek_X can only provided along
with a seek_X.

In the Python API, this verification is done dynamically when a user
message iterator class is assigned to a source or filter component
class.

Change-Id: If596d35dc3327bfd6e3f1e59f74c43dce3a722e1
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2100
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib: make can_seek_ns_from_origin logic use `can_seek_forward` property of iterator

This patch changes the behavior of
bt_self_component_port_input_message_iterator_can_seek_ns_from_origin to
make use of the new `can_seek_forward` property of iterators.

The current iterator "can seek ns from origin" logic works like this:

- If the iterator provides a `can_seek_ns_from_origin` method, call it
  and return that result.
- Otherwise, check if we can autoseek: if the iterator can seek
  beginning return true, else return false.

An issue with this is that if:

- an iterator doesn't provide the can_seek_ns_from_origin and
  seek_ns_from_origin methods
- the streams (and thus event messages) produced by this iterator don't
  have a timestamp
- the iterator can seek beginning

then
bt_self_component_port_input_message_iterator_can_seek_ns_from_origin
will report that this iterator can seek ns from origin, presumably
because it can use autoseek.  However, since the event messages don't
have a timestamp, we'll be able to seek beginning, but fast forwarding
to the desired point won't work.

To solve this, we need an additional property on message iterators,
whether the messages it produces have a timestamp, and therefore if the
"fast-forward" portion of autoseek will work.

This patch also brings another behavior change.  Previously, if the
can_seek_ns_from_origin method of the iterator was provided and returned
false, that was it.  With this patch, if that happens, we fall back to
checking if autoseek is supported.  That change is primarily in
bt_self_component_port_input_message_iterator_can_seek_ns_from_origin.
However,
bt_self_component_port_input_message_iterator_seek_ns_from_origin is
also updated accordingly, because we need to query the iterator to know
if it can seek ns from origin by itself, or if we should fall back on
autoseek.

Change-Id: I1848c87acf8ed75d4020a51e3be41dec2f144843
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2066
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

tests: remove CR characters from expected file in test_live

Tests done by test_compare_to_ctf_fs in test_live don't work on Windows.
When comparing the expected and actual files, the expected file has CRLF
end of lines while the actual file has LF end of lines.  This can be
unexpected, as both are produced by sink.text.details.

The reason is that bt_diff strips CR characters from the actual file.
This is done so actual files match the expect files, which typically are
static (not generated) and don't contain CR characters.

This patch makes test_live also strip CR characters from the expected
files it generates.

One might wonder, why not just make bt_diff strip both files of their CR
characters.  I prefer not to do that, because most files passed as
"expect" files to that function are files from the source tree.  These
files should not be modified by building and testing.  In other words,
it should be possible to run a make check while the source tree is on a
read-only filesystem.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: Ie5fac78260e5bb8102d3c7f9cc3b4a1b21cf773d
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2114
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>

tests: use os.rename instead of os.replace in lttng_live_server.py

On Windows, os.rename fails if the destination file exists, which is the
case for the port file in lttng_live_server.py.  The file is created by
the bash test script.

This results in the following error:

    # Starting LTTng live server mockup
    Traceback (most recent call last):
      File "C:/Users/smarchi/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1449, in <module>
        args.port_filename, args.sessions, args.max_query_data_response_size
      File "C:/Users/smarchi/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1229, in __init__
        self._write_port_to_file(port_filename)
      File "C:/Users/smarchi/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1342, in _write_port_to_file
        os.rename(tmp_port_file.name, port_filename)
    FileExistsError: [WinError 183] Cannot create a file when that file already exists: 'C:/Users/smarchi/AppData/Local/Temp/tmpi8ihb0ck' -> 'test_live_list_sessions_server_port.5iod4m'

Use os.replace instead.

Change-Id: I8d7a28da66f11975bb0e2d355084a8a336500cc0
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2113
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>

tests: convert paths passed to lttng_live_server.py

On Windows, the MSYS2 bash tries to convert paths in the Unix form
(/c/foo/bar) to the Windows form (C:\foo\bar).  However, some paths are
embedded in some complex arguments we pass to lttng_live_server.py, and
bash does not convert them.  This causes the file open in Python to
fail, because it doesn't find that file.  We therefore need to convert
them by hand using cygpath before passing them to the Python script.

Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Change-Id: I108e6dc2a690042bf727335fc5af5294bf928590
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2112
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

tests: name temporary files in test_live

I find it nice for debugging to have file names that have a little bit
of meaning.

Change-Id: I2505d5b363c2ea5f0c62255585e8898592d5ead0
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2111
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

Fix: ctf: open metadata file using mode "rb"

In the handling of the support info query, we open the metadata file
using the "r" mode.  On Windows, this is a problem when opening
packetized metadata files, which may contain bytes that Windows will
treat as end of lines and transorm.  Use "rb" instead.

This makes the test_trace_copy test pass on Windows.  It currently fails
with the "session-rotation" trace.

Change-Id: Iaa7a3b8def1c07335c5e7e094d82121df56de268
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2110
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Michael Jeanson <mjeanson@efficios.com>

cli: print map value in ASCII-betical order of keys

The test plugins/src.ctf.fs/query/test_query_metadata_info (and maybe
others that I don't know about) relies on the CLI outputting query
results in a constant way.  However, it isn't currently the case, as the
map entries are printed in the order that
bt_value_map_foreach_entry_const, therefore g_hash_table_iter_next,
provides them.  This test currently fails on the CI Windows machine for
this reason.

This patch makes the CLI print map values in a sorted order (defined by
comparing keys with strcmp).  This makes the tests stable, but it is
also convenient for the user, as the results will be sorted and in a
stable order.

Change-Id: I034b295bc238e146337844ac5d2c43a305fb90e0
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2109
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

tests: use bt_cli in test_trace_copy

This helps, by printing the executed babeltrace2 command lines.

Change-Id: Idaedfb212bd6cdb80a6498cec0e7020626ae0695
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2108
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

port: mmap offset alignment on Windows

On Windows the mmap offset is not aligned to the page size but to the
allocator granularity which is usually bigger than the page size. Add a new
internal function to the mmap compat code to get the proper offset alignment
size for each platform.

See the "dwFileOffsetLow" section for more details :
https://docs.microsoft.com/en-us/windows/win32/api/memoryapi/nf-memoryapi-mapviewoffile

Change-Id: Ic26dc53862719892738a88b3dcea6a15901a0c94
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2107
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

tests: delete temporary files in cli/test_help

Also, use `-t` option of mktemp so that temporary files are created in
the system's tmp directory (/tmp) instead of the current working
directory.

Change-Id: If27b7cc420a70033abc81f8f901521ea58a830bc
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2106
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Cleanup: tests: remove debugging `diag` leftover

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I6f537025ba758a06e730b752fc1298ff724352a9
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2105
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Rename `BT_PLUGIN_EXIT` to `BT_PLUGIN_FINALIZE`

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I5ac769f400d27688d97235db15d89a6c1c6e1ecb
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2104
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Rename `init` methods to `initialize`

We use the full word 'finalize' so why not use the full word
'initialize'.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I741e2406bf53a8da4b74a5335afcd7ec359727ea
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2098
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: split real FC/field into single and double prec FC/field

This commit removes the BT_FIELD_CLASS_TYPE_REAL field class and the
corresponding fields and adds BT_FIELD_CLASS_TYPE_SINGLE_PRECISION_REAL
and BT_FIELD_CLASS_TYPE_DOUBLE_PRECISION_REAL field classes and their
fields.

This seems like cleaner way to express the different types of floating
point representations and while keeping the door open for other
(smaller and/or larger) representations.

The test_field.py testcases for `_SinglePrecisionRealField` are
implemented using a default value of 52.0 which can be converted from
Python float(double precision) to C float(single precision) without loss
of precision. It's the best way I found to support the
`_TestNumericField` test cases without scattering specific conditions
for `_SinglePrecisionRealField`

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I6057e1626b71fee5dce1c374bda3faf3451b5fef
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2099
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: cli: don't log error when using help command with a plugin name

When using the help command with a plugin name, it works, but Babeltrace
logs an error:

    09-26 16:37:03.782 10054 10054 E CLI/CFG-CLI-ARGS plugin_comp_cls_names@babeltrace2-cfg-cli-args.c:189 Missing component class type (`source`, `filter`, or `sink`).

This is problematic, because this is a valid use of the command.
Logging an error is unnecessarily alarming.

The current code takes the argument (in our case, a plugin name such as
"ctf") and passes it through plugin_comp_cls_names.  The function can't
parse the string as a component class name, so logs the error mentioned
above.  The caller, bt_config_help_from_args, sees that the function has
failed, so falls back to assuming the string is a plugin name, which
then works correctly.

To avoid this, the sophisticated solution I chose was to look for an
unescaped dot in the argument.  If there's one, we treat it as a
component class name, if there isn't, we treat is as a plugin name.

Change-Id: I76b791e5d44436fdc5b10a3c10e3579a0783b04d
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2101
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

Fix: bt2: duplicate test name in test_field.py

Flake8 reports:

./tests/bindings/python/bt2/test_field.py:2170:5: F811 redefinition of unused 'test_has_field_prop_true' from line 2166
./tests/bindings/python/bt2/test_field.py:2178:5: F811 redefinition of unused 'test_bool_op_true' from line 2174

I think that in both cases, the second instance was meant to be renamed
to `_false`.

Reported-by: flake8
Change-Id: I8b3143bb3f86982e220a6b516f028fca9e6e24ab
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2093

bt2: remove unused variable in UserMessageIteratorSeekNsFromOriginTestCase.test_seek_ns_from_origin

Flake8 reports:

./tests/bindings/python/bt2/test_message_iterator.py:790:9: F841 local variable 'msg' is assigned to but never used

Reported-by: flake8
Change-Id: I4be9bf2d1a86e4c5d538fd1d1461f6e2bbf2b763
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2092
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Fix: bt2: duplicate test name in _IntegerRangeTestCase

Flake8 reports:

./tests/bindings/python/bt2/test_integer_range_set.py:45:5: F811 redefinition of unused 'test_create_wrong_type_lower' from line 41

I think the issue is simply that the second instance should have `upper`
instead of `lower`.

Reported-by: flake8
Change-Id: Id03b801951b9ce64eb0264debdd8ed4c4e3a577e
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2091

Fix: bt2: use of undefined variable in _VariantFieldClassWithSelectorTestCase.test_const_append

Flake8 reports:

./tests/bindings/python/bt2/test_field_class.py:875:43: F821 undefined name 'str_field_class'

Fix it by defining the str_field_class variable (even though it won't
actually be used).

Reported-by: flake8
Change-Id: If17cef30ed24b6ec9d1ee25d3f6c94d491ffd8fd
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2090

Fix: bt2: unused assignments in test_field_class.py

Fix these:

    ./tests/bindings/python/bt2/test_field_class.py:101:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:124:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:215:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:234:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:251:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:402:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:428:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:445:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:1183:9: F841 local variable 'field' is assigned to but never used
    ./tests/bindings/python/bt2/test_field_class.py:1222:9: F841 local variable 'field' is assigned to but never used

In all these cases, the assignment has no effect.  The original intent
was probably to do `field.value = ...`, so use that everywhere.

Reported-by: flake8
Change-Id: If28fab3d1a8facb0123ab3a2dc0b83062ebed38e
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2089

bt2: print actual and expected lengths in _StaticArrayField._set_value error message

I think it can help the user figure out what's wrong if we give the
lengths of the field and the passed value in the error message.

Change-Id: I94fb7ca04724d5d7af701df88aaedc195b1736fd
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2088
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: ignore comparison to None flake8 warnings in test_value.py

Fix these:

    ./tests/bindings/python/bt2/test_value.py:485:36: E711 comparison to None should be 'if cond is None:'
    ./tests/bindings/python/bt2/test_value.py:488:35: E711 comparison to None should be 'if cond is not None:'
    ./tests/bindings/python/bt2/test_value.py:1177:36: E711 comparison to None should be 'if cond is None:'
    ./tests/bindings/python/bt2/test_value.py:1180:35: E711 comparison to None should be 'if cond is not None:'

In these tests, we specifically want to test that comparing (equal/not
equal) to None yields the expected result.

Reported-by: flake8
Change-Id: I47d96598a158b10116fc6a948d42ef61f608fb07
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2087

bt2: remove duplicate test test_create_from_vint in RealValueTestCase

There are two tests named like this in RealValueTestCase.  The second
one looks right (it creates a Value from an int, then creates a
RealValue from that).  The first one looks like it should be called
"test_create_from_vfloat", as it tests creating a RealValue from another
RealValue, but we already have such a test case a bit lower.  So I think
it's fine to delete it.

Reported-by: flake8
Change-Id: I274b922c653675913f2346cbc67cbcd03cd5208f
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2086
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: remove duplicated test test_setitem_none in ArrayValueTestCase

The two tests with that name are identical, so remove one.

Reported-by: flake8
Change-Id: Ibb63b6f548bceb81d6173dcc9619978e62d8828e
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2085
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: rename test with conflicting name in test_trace.py

The second test named `test_env_iter` seems to be testing iteration of a
const environment object, so rename it to `test_const_env_iter`.

Reported-by: flake8
Change-Id: I614756daa511b9c66b031328f11ad77053ab9fc5
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2084
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Add setup.cfg with sensible flake8 configuration

This is a sensible config which allows to run "flake8 ." without having
too many useless warnings.

Note that the `per-file-ignores` config requires flake8 >= 3.7.

Ignore:

- E501, line too long, because we don't really care about that (black
  decides how it's formatted anyway).
- W503, line break before bbinary operator, because we don't really care
  about that (black decides how it's formatted anyway).

__init__.py contains some expectedly unused imports, so it seems cleaner
to disable that warning for the whole file, rather than putting a "noqa"
comment on each line.  Of course, we risk having a really unused import
in that file, but it doesn't contain much, so the risk is low.

Finally, exclude the Python tap directory, as it's not our code.

Change-Id: Iebdf274b91907fb31d4ebfa6ae4c8157de46467c
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2083
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib: pass config object to message iterator init method, add can seek forward property

This patch introduces a new bt_self_message_iterator_configuration
object that is passed to message iterator init methods.  The purpose of
this object is to hold options the message iterator can set at init time
but not later.

To verify that the user does not save a pointer to this object and try
to use it later, the object is frozen right after the init method
returns.

The object contains one option that we know we'll need,
`can_seek_forward`.  The corresponding functions to set it (on the
config object) and read it (from the self component port input message
iterator object) are also provided.  But because this patch already
contains a lot of boilerplate, the code that actually uses that option
will come in a following patch.

The Python bindings provide a simple wrapper around this new object.

Change-Id: Ibcd0df842fd154d3bafe3869b56f70aa3a72ceab
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2064
Tested-by: jenkins <jenkins@lttng.org>

Fix: tests: live: listen on python < 3.5 needs backlog parameter

The backlog parameter is required on python version smaller then 3.5.

On sles12sp2:

Traceback (most recent call last):
   File "/home/jenkins/workspace/babeltrace_master_slesbuild/arch/sles12sp2/build/std/conf/prod/src/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1447, in <module>
     args.port_filename, args.sessions, args.max_query_data_response_size
   File "/home/jenkins/workspace/babeltrace_master_slesbuild/arch/sles12sp2/build/std/conf/prod/src/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1232, in __init__
     self._listen()
   File "/home/jenkins/workspace/babeltrace_master_slesbuild/arch/sles12sp2/build/std/conf/prod/src/babeltrace/tests/data/plugins/src.ctf.lttng-live/lttng_live_server.py", line 1323, in _listen
     self._sock.listen()
TypeError: listen() takes exactly one argument (0 given)

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: Ib923beb9eadf929714b3270aaa7a57ff6c52407e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2096
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

lib: pass config objects to component init methods

This patch introduces configuration objects that are passed to init
methods of components.  The goal is to make it possible to have
parameters that can be set during the component's init phase, and not
after.

At the moment, they are not used for anything, they are only there for
future use.  For that reason, we currently just pass the value NULL as
the config objects to the components.  This should be fine, since the
components should not do anything with the config objects for now.

Similarly, we have configuration Python objects passed to __init__
methods of _UserMessageIterator objects.  These configuration objects
are empty, but we can connect them with the corresponding SWIG pointer
once we actually add a method in them.

I added tests to verify that the config objects the components receive
have the right types.

Change-Id: If7568160e225c1eb8343eb212f8b4961f5472836
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2063
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: make it possible to remove a trace class or trace destruction listener

This patch adds the possibility to remove a previously added trace or
trace class destruction listener. Calling
{_TraceClass,_Trace}.add_destruction_listener returns a
_ListenerHandle. To remove the listener, the user must pass this handle
back to {_TraceClass,_Trace}.remove_destruction_listener.

Change-Id: I6020135c1e056658d941d2e022b171466c5df7b3
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2077
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: make it work for Python 3.4

The import system from Python 3.4 has some significant differences
compared to the one from Python 3.5 and later.  Notably, support for
circular references (which we use) was added in the later versions.

For example, trace.py and stream.py both import each other.  Trace
objects need Stream objects when listing streams in a trace, and Stream
objects need Trace objects when getting the containing trace of a
stream.

To make it work on Python 3.4, I think we need to break any circular
import chain that happen when "import bt2" is done.  It is fine to
import things later, once all the modules have been initialized.

To achieve this, I changed some

  from bt2 import foo as bt2_foo

to

  def _get_bt2_foo():
      from bt2 import foo as bt2_foo
      return bt2_foo

The users of bt2_foo in that file now need to use _get_bt2_foo() instead
of bt2_foo.  it is important that _get_bt2_foo doesn't get called during
the import of the module, otherwise, we come back to the initial
problem.  So in some cases, I have wrapped some calls to _get_bt2_foo in
properties, as was already done to break circular dependencies between
classes.

This is not enough though, as there is another behavior difference
between 3.4 and 3.5+, which I believe is also a consequence of the
circular import changes [1].  It is also related to the fact that we
delete submodule attributes from the bt2 module in our __init__.py.

With 3.5, it is still possible to do:

    from bt2 import value

after we've deleted the `value` attribute from bt2.  With Python 3.4, it
doesn't work.  I believe it is related to [1] in that Python 3.5+, when
processing a "from" import, searches in sys.modules if a simple getattr
on the module failed.

This is problematic for us, since we need the 'from bt2 import "foo"'
lines to work after the bt2 attributes have been deleted (to support the
changes described above).  The solution I've found is to not delete the
bt2 submodule attributes when using Python 3.4.

We delete those attributes simply to clean up the bt2 namespace and to
avoid exposing the internal organisation of the code.  I think it's
an acceptable trade-off to not do it when using Python 3.4, if it means
that it will actually work.

[1] https://bugs.python.org/issue17636

Change-Id: Ia6810972492a058f60c21e6f22afd43962f4f7a2
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2082
Tested-by: jenkins <jenkins@lttng.org>

Update project plugins's properties

This patch updates various properties of the project's plugins. Those
properties are publicly visible through the API or with the CLI's `help`
command.

The updated properties are:

Descriptions:
    Updated to add details or normalize them.

Authors:
    All changed to `EfficiOS <https://www.efficios.com/>`. Keeping a
    specific up-to-date list of authors will be challenging in the
    future, especially with many contributors.

Help texts:
    Added to refer to specific manual pages.

    Maybe we can find a way to set a manual page's plain text as a
    component class's help text in the future, but in the meantime at
    least the user knows where to look for help.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I44d13a89039e433c06e1c527eb2be577fd471454
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2047
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

utils.sh: expand `${extra_details_args[@]}` conditionally

When you call bt_diff_details_ctf_single() or
bt_diff_details_ctf_gen_single(), it is possible that you don't provide
any extra `sink.text.details` arguments. This makes
`${extra_details_args[@]}` empty.

`utils.sh` does

    set -u

to make Bash throw an error when a variable is undefined/unbound.

Starting from Bash 4.4 [1]:

> Using ${a[@]} or ${a[*]} with an array without any assigned elements
> when the nounset option is enabled no longer throws an unbound
> variable error.

In bt_diff_details_ctf_single() and bt_diff_details_ctf_gen_single(), we
use `"${extra_details_args[@]}"` directly: Bash < 4.4 throws an error if
the array is empty.

We don't get this error on CI nodes having Bash < 4.4 currently because
all the usages of bt_diff_details_ctf_single() and
bt_diff_details_ctf_gen_single() pass extra `sink.text.details`
arguments. However it is my understanding that they are optional.

This patch replaces

    "${extra_details_args[@]}"

with

    ${extra_details_args[@]+"${extra_details_args[@]}"}

so that `"${extra_details_args[@]}"` is only used when
`${extra_details_args[@]}` is defined. This works with any Bash version.

[1]: https://git.savannah.gnu.org/cgit/bash.git/tree/CHANGES?id=3ba697465bc74fab513a26dea700cc82e9f4724e#n878

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: Ic74117607ac616f3e27968a6f8fb09d07fe7eb12
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2080
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

sink.text.details: add `with-data` initialization parameter

This patch makes a `sink.text.details` component handle a new
`with-data` initialization parameter. This new parameter controls
whether or not the component writes data lines, as opposed to metadata
lines: anything that's not metadata is considered data, except for
message iterator inactivity messages.

This can be useful to inspect the metadata blocks of the received
messages without having all the message lines.

The new `default-without-data-without-metadata.expect` expectation file
is empty because, without data and without metadata, there's effectively
nothing to write.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I844073b2080df373501a9216d8041f4d6b799c08
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2076
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Add basic `sink.text.details` tests

This patch adds very basic `sink.text.details` tests to at least keep
track of the generated text's format. The tests test each initialization
parameter individually, and some combined with `compact=yes`.

As of this patch, all tests pass `with-stream-name=no` to initialize the
`sink.text.details` component because the `src.ctf.fs` component uses
absolute paths to set stream names and this is not compatible with
static expectation files.

Future tests should use a custom source to check how `sink.text.details`
writes the stream name property.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I79c4956e5d67dd922130443992de9daef2974f4e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2075
CI-Build: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

sink.text.details: don't write `:` after `Packet beginning` without ctx

This patch fixes packet beginning lines which look like:

    ...
        Stream (ID 0, Class ID 0)

    {Trace 0, Stream class ID 0, Stream ID 0}
    Packet beginning:

    {Trace 0, Stream class ID 0, Stream ID 0}
    ...

so that they become:

    ...
        Stream (ID 0, Class ID 0)

    {Trace 0, Stream class ID 0, Stream ID 0}
    Packet beginning

    {Trace 0, Stream class ID 0, Stream ID 0}
    ...

In other words, the `:` character following `Packet beginning` is
useless as there's no (packet context) data following.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I4c75bc487c10fa2a4b39708d09d4089887ceabbd
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2074
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

src.ctf.fs: deterministically sort trace's DS file groups

This patch makes a `src.ctf.fs` component deterministically sort its
trace's DS file groups. The order of the DS file group array directly
influences:

* The order of the output ports.
* Which DS file group gets which stream ID if the trace's packet headers
do not contain a `stream_instance_id` field (automatic stream IDs).

It's important that, whatever the build or file system, the component
always assigns the same IDs to the same DS file group for a given trace.
This can help debugging and testing.

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I105f44e7885a2974ace4d62abe0ee5384316b350
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2079
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Tests: src.ctf.lttng-live: add inverse ordering from relayd

Validate that relayd side ordering does not have an influence on
determinism of trace decoding.

Change-Id: I37cf7f0d17616acb667339793d6624a7c67dfae3
Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1916
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Add LTTng-independent `src.ctf.lttng-live` tests

This patch adds tests for `src.ctf.lttng-live`. It serves as a foundation
for future `src.ctf.lttng-live` tests.

Because `src.ctf.lttng-live` needs a running LTTng relay daemon to
connect to and because we don't want the Babeltrace tests to depend on
LTTng, the new `lttng_live_server.py` script mimics an LTTng relay
daemon. The script expects exact commands from an LTTng live viewer and
replies with the expected responses. It is mainly a scoped down
lttgn-relayd.

The faux server script accepts multiple arguments:

positional arguments:
  SESSION (for each session)
    A session configuration. There is no space after
    comma.
    Format is:
    'NAME,ID,HOSTNAME,FREQ,CLIENTS,TRACEPATH[,TRACEPATH]....'

optional arguments:
  --log-level {info,warning} (default: warning)
    The loglevel to be used for the server.

  --port-filename PORT_FILENAME (required)
    The final port file. This file is present when the server is ready to
    receive connection. This is used to as a synchronization point for
    the shell test script.

  --max-query-data-response-size MAX_QUERY_DATA_RESPONSE_SIZE (optional)
    The maximum size of control data response in bytes. This is used to force
    the client to perform multiple requests for the same packet.

The faux server bind to port 0 and get assigned a free port. This prevent
unforeseen interaction with the outside world

In the test script, `tests/plugins/src.ctf.lttng-live/test_live`, the
execution goes as follow:

1. It runs the faux server, passing two temporary file names, and making
   it a background job.

2. It waits for the second temporary file to contain anything.

3. It reads the port number from the second temporary file.

4. It runs:

       $ babeltrace2 -i lttng-live net://localhost:PORT/[valid path]

   where `PORT` is the faux server's port number.

5. It compares the output of 4. to an expected file generated from a
   template.

6. It removes the temporary files.

The faux server exits itself.

The current tests are:
  - test_list_sessions
    List 2 sessions including one with traces for multiple domains.
  - test_base
    Attach to the live server and read a standard ust trace with no discarded
    events.
  - test_multi_domains
    Attach and consume a multi-domain trace with discarded
    events.
  - test_rate_limited
    Attach and consume data from a multi packets ust session with no
    discarded events. Enforce a server side limit on the stream data
    requests size.
  - test_compare_to_ctf_fs
    Attach and consume a multi-domain trace with discarded
    events and compare it to the same trace read by src.ctf.fs.

Missing tests mostly concerns multiple traces under a same session
(per-pid) on listing and reading.

This test only runs if you have a Python interpreter (`HAVE_PYTHON`
Automake definition).

Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien@efficios.com>
Change-Id: I2919de3cdfa06eda6f4fcf118689571c13358246
Reviewed-on: https://review.lttng.org/c/babeltrace/+/1915
CI-Build: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

bt2: field.py: raise ValueError when setting out of range value to Integer

This commit removes the use of `utils._check_int64()` and
`utils._check_uint64()` methods as the validity of the value is now
checked using the _check_range() method and the existing field class'
`_field_value_range` setter property.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ib5b33ae5becc9f71b20b17ae0244b6e676b89f75
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2081
Tested-by: jenkins <jenkins@lttng.org>

bt2: remove unused import

trace_class.py doesn't need to import trace_class.

Change-Id: I792b742a6aecd9903f64815e4cf066d05d1ccc24
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2078
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Fix: muxer_upstream_msg_iters_can_all_seek_beginning(): init. `status`

Detected by GCC:

    muxer.c: In function ‘muxer_msg_iter_can_seek_beginning’:
    muxer.c:1541:9: error: ‘status’ may be used uninitialized in this
    function [-Werror=maybe-uninitialized]
      return status;
             ^~~~~~
    muxer.c:1529:5: error: ‘status’ may be used uninitialized in this
    function [-Werror=maybe-uninitialized]
      if (status != BT_COMPONENT_CLASS_MESSAGE_ITERATOR_CAN_SEEK_BEGINNING_METHOD_STATUS_OK) {
         ^

Signed-off-by: Philippe Proulx <eeppeliteloop@gmail.com>
Change-Id: I55ad73aeb9595796bdcddff5ba4d2895d5f2314e
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2073
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

flt.utils.muxer: remove `assume-absolute-clock-classes` param

It's a now a postcondition that all the messages of a given message
iterator have compatible clock classes. Thus the
`assume-absolute-clock-classes` parameter is useless as there is no more
need to force a muxer to accept possibly non-compatible clock classes
because we check that they are indeed all compatible.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I1efcd99abfd1da1d9a4b7c9c6962f898446d31d1
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2067
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Jonathan Rajotte Julien <jonathan.rajotte-julien@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

cli: make `--clock-force-correlate` use `force-clock-class-origin-unix-epoch`

The flt.utils.muxer's `assume-absolute-clock-classes` parameter will be
removed in a subsequent commit.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I282576ad778eefd40e1d57266f3af2a880ceca47
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2071
Tested-by: jenkins <jenkins@lttng.org>

src.ctf.fs: add `force-clock-class-origin-unix-epoch` param

The `force-clock-class-origin-unix-epoch` initialization parameter makes
all the created clock classes have a Unix epoch origin. This is useful
to force the clock classes of multiple traces to be compatible even if
they are not inherently.

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: If311081052e9be28ca5782ee684b4ef7bba66176
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2069
Tested-by: jenkins <jenkins@lttng.org>

bt2: print error causes in bt2._Error.__str__

When working on Python tests where a bt2._Error escapes the test, I very
often write the same thing to be able to visualize error causes:

    try:
        ... original code ...
    except bt2._Error as e:
        for c in e:
    print(c)
raise

By making bt2._Error.__str__ print the error causes, the causes will be
printed automatically when a test fails this way.

And from the point of view of a user of the bt2 package, I think it's
quite useful to get the error causes right away, rather than having to
write a snippet like the above.  As a user of the library, if a
bt2._Error escapes or if I print it, I'd be happy to get this very
precise information.  Also, not everyone might know from the start that
it's possible to print detailed causes, and the otherwise default
message is rather vague.  It just says for example "Failed to add
component to graph", without any other detail.  So I think we would do
users a service by printing the causes by default.

Change-Id: I6f8c72b8e358d6e664c47a1a43c9a81c91a76239
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2070
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

tests: utils.sh: add lttng-utils plugin to plugin path

Change-Id: I41983a0c42ea2e41b385110e5e8ea003f9b24b3c
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2072
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

Cleanup: clock-snapshot.c: logically dead code

Coverity reported the following:
CID 1401189 (#1 of 1): Logically dead code (DEADCODE)dead_error_begin:
Execution cannot reach this statement: bt_clock_snapshot_recycle(c.

Reported-by: Coverity - 1401189 Logically dead code (DEADCODE)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I382a29cf5d825a9df06de61cbed7dcdf988de7d3
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2053
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: test_dwarf.c: close() argument cannot be negative

Coverity reported the following:
CID 1401801 (#1 of 1): Argument cannot be negative (NEGATIVE_RETURNS)8.
negative_returns: fd is passed to a parameter that cannot be negative.

Reported-by: Coverity - Argument cannot be negative (NEGATIVE_RETURNS)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ibcdcfd54f5f711923e5a31544c86e82da3ec6226
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2048
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Cleanup: ctf-writer: field-wrapper.c: logically dead code

Coverity reported the following:
CID 1401191 (#1 of 1): Logically dead code (DEADCODE)dead_error_begin:
Execution cannot reach this statement: bt_ctf_field_wrapper_destroy.

Reported-by: Coverity - 1401191 Logically dead code (DEADCODE)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ib468f8df4a11491b185ffa45c2db3bc4fc8fb045
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2051
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: ctf-writer: stream-class.c: Dereference after null check

Coverity reported the following:
  CID 1401231 (#1 of 1): Dereference after null check (FORWARD_NULL)8.
  var_deref_model: Passing null pointer ft_copy to
  bt_ctf_field_type_common_integer_set_mapped_clock_class_no_check_frozen,
  which dereferences it

Reported-by: Coverity - 1401231 Dereference after null check (FORWARD_NULL)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I32a15f0ff927c8b987e785a37596e9c56b628de5
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2049
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: flt.lttng-utils.debug-info: bin-info.c: Dereferencing null pointer

Coverity reported the following:
CID 1401211 (#1 of 1): Dereference after null check (FORWARD_NULL)4.
var_deref_op: Dereferencing null pointer bin.

Reported-by: Coverity - 1401211 Dereference after null check (FORWARD_NULL)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: I66c3dc42305b1c14a5567141274249c6e6b10082
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2050
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: object-pool.c: dereferenced before null check

Coverity reported the following:
  CID 1401187 (#1 of 1): Dereference before null check
  (REVERSE_INULL)check_after_deref: Null-checking pool suggests that it
  may be null, but it has already been dereferenced on all paths leading
  to the check.

Add a `BT_ASSERT()` to ensure that `pool` is not null.

Reported-by: Coverity - 1401187 Dereference before null check (REVERSE_INULL)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ie18fecdad17c1a42db914d2eb987f783a4523871
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2054
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Cleanup: packet-context-field.c: logically dead code

Coverity reported the following:
CID 1401199 (#1 of 1): Logically dead code (DEADCODE)dead_error_begin:
Execution cannot reach this statement: bt_field_wrapper_destroy

Reported-by: Coverity - 1401199 Logically dead code (DEADCODE)
Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Change-Id: Ia53f46f998bcfb3a88f2d8a93343aa4c199a12ff
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2052
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>

Fix: output non-LTTng CTF trace with same relative path as input

Commit

  73760435c5a4 ("cli: automatically detect sources for leftover arguments")

changed the behavior of the CLI when converting non-LTTng CTF traces
[1].  The intended behavior, prior to this commit, is to replicate in
the output directory the directory structure from the input directory.

So let's say we have CTF traces at `a/b/c1` and `a/b/c2` (metadata files
inside those directories) and we do:

    babeltrace2 a -c sink.ctf.fs -p 'path="out"'

We expect to have traces output in `out/b/c1` and `out/b/c2`.

Before the introduction of automatic source discovery, the src.ctf.fs
component would receive `a` as the input path, recurse to find the
traces, and set the name of each trace to be the path of the trace
relative to `a` (`b/c1` and `b/c2` in the example above).  The
sink.ctf.fs component would then use add the trace name to the
user-provided output path to build the output path of each trace.

Now, when used with automatic source discovery (and especially since
they have lost the ability to recurse), src.ctf.fs components receive
the direct path to each trace (one trace per component).  So, in the
example above, that is `a/b/c1` and `a/b/c2`.  Since they don't have
access to the top-level directory the user specified, they can no longer
compute the trace name that the sink.ctf.fs component so desperately
needs.

Currently (since 48881202c2ef ("ctf: make src.ctf.fs not recurse")), the
normalized path is passed as the trace name.  So we end up with the
absolute path of the input trace under the output directory.

To restore the original behavior, this patch makes the CLI (which has
access to the required information)  compute the trace name the same way
src.ctf.fs used to do, and pass it as the `trace-name` parameter to
src.ctf.fs components.  It only does so when the trace has a single
input directory.

[1] This behavior only affects non-LTTng CTF traces, because sink.ctf.fs
first tries to build the output path using certain environment fields
(which LTTng traces provide), and only falls back on the trace name if
the required fields are not present.  So for LTTng traces (or other
traces that have the required env fields), the trace name does not
matter.

Change-Id: Iaba02fbae856ac5ddc11724613457bcc0b5b4019
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2026
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

cli: don't include version.h in babeltrace2-cfg-cli-params-arg.c

The only thing provided by version.h is the GIT_VERSION macro, which
this file doesn't use. Because it includes that header file, it is
unnecessarily rebuilt after we do or amend a git commit.

Change-Id: Ic7526f09f62896d5dfc299f15205fe9982060c4d
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2065
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: add `if __name__ == '__main__'` snippet to all tests

I find it useful to run a single test file.

Change-Id: I6387864abf246977b964e17b932beebdba8bc92c
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2062
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: remove native_bt.d in `make clean`

Change-Id: Ifedefce4d1d0f7ffad2fa9f32cb59d581750bb46
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2061
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

Fix: ctf: assert that name is not NULL in warn_meaningless_field()

gcc 9.1 reports that warn_meaningless_field's `name` is sometimes
NULL when it is inlined in warn_meaningless_fields(), itself
inlined in ctf_trace_class_warn_meaningless_header_fields() at
line 131.

The the only call site using a `NULL` name will have
`fc->type == CTF_FIELD_CLASS_TYPE_STRUCT`.

Add a `BT_ASSERT(name)` to warn_meaningless_field() and a comment
indicating that `name` is guaranteed to be non-NULL whenever the field
class is not a structure.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I323b9f390f80dbd898a69664082fd07d5cc2fc86
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2000
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

ctf: silence bogus warning in set_field_refs()

gcc 9.1.0 warns that struct_fc my be used uninitialized. The code
flow doesn't allow this to happen, but this warning is annoying/scary
and easily fixed by initializing struct_fc to NULL.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: Ib4c702dfc0ca4a29e387b384a1603bdf4634c599
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2028
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

ctf: silence bogus warnings in create_relative_field_ref()

gcc 9.1.0 warns that struct_fc and var_fc may be used uninitialized.
The code flow doesn't allow this to happen, but this warning is
annoying/scary and easily fixed by initializing both variables to
NULL.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I554affb70275f60eff08134f80b33bc4cefbc91f
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2027
Reviewed-by: Simon Marchi <simon.marchi@efficios.com>
Tested-by: jenkins <jenkins@lttng.org>

bt2: expose seek_ns_from_origin and can_seek_ns_from_origin

This patch makes the seek_ns_from_origin and can_seek_ns_from_origin
operations accessible in Python.  That means:

- A message iterator class implemented in Python can define the
  _user_seek_ns_from_origin and _user_can_seek_ns_from_origin
  methods to implement these operations.
- A message iterator class implemented in Python that has an upstream
  message iterator can call these operations on it using methods on that
  upstream message iterator object.

There is no big surprise in the patch, as everything has mostly been
modeled on seek_beginning/can_seek_beginning.

Change-Id: Ic726b64b4d7c24a59bab03df8a77cf8369bb5e5f
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2020
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

bt2: test_message_iterator: use assertIs instead of assertTrue/assertFalse in can seek tests

In these tests, we want to specifically check that the return values are
True or False, not any value that evaluates to True or False.

For example, we initialize can_seek_beginning to None, run
graph.run_once(), then do assertFalse(can_seek_beginning). If, for some
reason, can_seek_beginning is still None, the test will wrongfully pass.

The same could happen if can_seek_beginning gets assigned some value
that evaluates to True (other than True itself) and we use
assertTrue(can_seek_beginning).

Instead, use assertIs, which checks that the value is specifically the
one and only True or the one and only False.

Change-Id: Idf94d350ecc06cdb6a38ca31b2d77804fbf57df7
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2019
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: remove _GenericMessageIterator abstraction level

We used to have two message iterator types sharing a lot,
_OutputPortMessageIterator and _UserComponentInputPortMessageIterator,
so they shared some code in the name of _GenericMessageIterator.
_OutputPortMessageIterator is no more, so the _GenericMessageIterator
abstraction is not useful anymore. Rename to
_UserComponentInputPortMessageIterator and remove all the templating.

Change-Id: If626aeac881ab2ca478a196a53047b4bcf3b7a65
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2018
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

bt2: make can_seek_beginning a method instead of a property

_GenericMessageIterator.can_seek_beginning and
_UserMessageIterator._user_can_seek_beginning are currently properties.
We will add the equivalent for can_seek_ns_from_origin, but this one
requires a parameter (the ns_from_origin value), so it can't be a
property, it has to be a method. For consistency, change the
can_seek_beginning ones to be methods.

Change-Id: Ic2c5b54908753dc7498fb1d0f94022301f766a15
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2017
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>

lib: remove END values from bt_message_iterator_seek_{beginning,ns_from_origin}_status

The corresponding status enumerations on the component class side:

- bt_component_class_message_iterator_seek_beginning_method_status
- bt_component_class_message_iterator_seek_ns_from_origin_method_status

don't include END, so it's not possible for
bt_self_component_port_input_message_iterator_seek_beginning or
bt_self_component_port_input_message_iterator_seek_ns_from_origin to
return END either.

Change-Id: I57a469d724f9605cd6a169459a39a7286f0e191b
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2015
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Reviewed-by: Philippe Proulx <eeppeliteloop@gmail.com>

lib: make can_seek_beginning and can_seek_ns_from_origin methods return a status

The current situation is problematic: calling any of these methods on a
message iterator can result in running some user code, which can fail
for various reasons (e.g. any Python exception being raised and
escaping).  It is not possible to inform the caller of such failure, as
the return type of these methods is bt_bool.  This must change.

This patch changes them to return a status, and return the result by an
output parameter.

If the method returns OK, it must have set the output parameter to
BT_TRUE or BT_FALSE.  If it returns another status, the output parameter
value is unused.

Change-Id: Iaa3dcc1494e2a0fcbdbb14283af3bf8504cd8ef5
Signed-off-by: Simon Marchi <simon.marchi@efficios.com>
Reviewed-on: https://review.lttng.org/c/babeltrace/+/2012
Tested-by: jenkins <jenkins@lttng.org>
Reviewed-by: Francis Deslauriers <francis.deslauriers@efficios.com>