[PING] [PATCH 0/3] Fix register_test_foreach_arch

[Lancelot SIX <Lancelot.Six@amd.com>]

Hi,

Kindly pinging.

Best,
Lancelot.

On 01/04/2022 14:00, Lancelot SIX wrote:
> Hi,
> 
> While working on a selftest issue, I noticed that there is a silent
> problem with register_test_foreach_arch:
> 
>          $ ./gdb/gdb \
>              -data-directory ./gdb/data-directory \
>              -quiet -batch -ex "maint selftest" 2>&1 \
>              | grep -E "Ran [0-9]+ unit tests"
>          Ran 145 unit tests, 0 failed
>          $ GDB_REVERSE_INIT_FUNCTIONS=1 ./gdb/gdb \
>              -data-directory ./gdb/data-directory \
>              -quiet -batch -ex "maint selftest" 2>&1 \
>              | grep -E "Ran [0-9]+ unit tests"
>          Ran 82 unit tests, 0 failed
> 
> Notice that when the GDB_REVERSE_INIT_FUNCTIONS env variable is set, the
> number of tests registered is different compared to when this env
> variable is not set.
> 
> This is due to an issue with how register_test_foreach_arch is used and
> how it works.  This function is used to register a generic selftest and
> run it against each architecture GDB supports.  Such registration is
> typically done during GDB's initialization (in _initialize_* functions).
> At the point where this function is called, it will iterate over all the
> architectures known to GDB and create an instance of the selftest for
> it.  The problem is that at that point, because GDB's initialization
> is not done yet, the best register_test_foreach_arch can do is register
> the test for each architecture initialized so far.  All architectures
> which are initialized at a later point will not have the selftest.
> 
> To fix this, this series proposes a lazy registration mechanism so
> register_test_foreach_arch can delay the moment it iterates over the
> known architectures to a point when GDB is fully initialized.
> 
> The first patch of the series does some preparatory refactoring which
> will come in handy later.  Patch 2 introduces the lazy test generation
> mechanism in the selftest framework, and patch 3 uses this mechanism to
> fix register_test_foreach_arch.
> 
> Tested on x86_64-gnu-linux.
> 
> All feedbacks are welcome.
> 
> Best,
> Lancelot.
> 
> Lancelot SIX (3):
>    gdbsupport/selftest: Replace for_each_selftest with an iterator_range
>    gdbsupport/selftest: Allow lazy registration
>    gdb/selftest-arch: Make register_test_foreach_arch generate arch tests
>      lazily
> 
>   gdb/maint.c                        | 13 ++++-----
>   gdb/selftest-arch.c                | 29 +++++++++++++------
>   gdb/selftest-arch.h                |  3 ++
>   gdb/testsuite/gdb.gdb/unittest.exp | 25 ++++++++++++++--
>   gdbsupport/selftest.cc             | 46 +++++++++++++++++++++---------
>   gdbsupport/selftest.h              | 42 ++++++++++++++++++++++-----
>   6 files changed, 120 insertions(+), 38 deletions(-)
> 
> 
> base-commit: 5530c021ce01abf368a9cde26b22c4b34f320ee8