Hi,
this looks kind of wrong to me:
1. If you speak about the nondeterministic nature of the addresses of
the events being different: That does not matter, the order of the event
subscriptions within the array of different events do not matter. Only
the order of the subscriptions of the same event do matter. This is
coming from the fact that we are only "using" the resulting order of all
subscriptions by its history, not by its address.
2. If you speak about the fact that we rely on the same events in the
array keeping the same order, and there is behavior in the windows qsort
implementation that is breaking our assertions, then this needs to be
fixed in the EFL_CALLBACKS_ARRAY_DEFINE macro, otherwise every single
user gets the wrong behavior. Which has (exactly in this spot) caused
apps to not even start in the past. (There was a discussion in the past
about the fact that qsort does not have documented defined behavior for
maintaining order of equal event, however, there is some black magic we
could apply to get that, with rather low memory/perf impact)
Greetings,
bu5hm4n
On 11/4/20 7:56 PM, Wander Lairson Costa wrote:
raster pushed a commit to branch master.
http://git.enlightenment.org/core/efl.git/commit/?id=fc949660f70f99fe97e63de65b1b67377d16c6da
commit fc949660f70f99fe97e63de65b1b67377d16c6da
Author: Wander Lairson Costa <wander.lair...@gmail.com>
Date: Wed Nov 4 18:25:37 2020 +0000
eo_test_general.c: Make eo_signals tests pass on Windows
Summary:
EFL_CALLBACKS_ARRAY_DEFINE reorders the callbacks according to
efl_callbacks_cmp. efl_callbacks_cmp compares the address of the desc
field, which depends on the memory layout generated by the linker.
To make the test run deterministically, we define the array of callbacks
manually.
Reviewers: vtorri, felipealmeida, raster
Reviewed By: raster
Subscribers: cedric, #reviewers, #committers, jptiz, felipealmeida
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D12043
---
src/tests/eo/suite/eo_test_general.c | 25 ++++++++++++++++++++-----
1 file changed, 20 insertions(+), 5 deletions(-)
diff --git a/src/tests/eo/suite/eo_test_general.c
b/src/tests/eo/suite/eo_test_general.c
index ae026a27f4..7bdb0e170b 100644
--- a/src/tests/eo/suite/eo_test_general.c
+++ b/src/tests/eo/suite/eo_test_general.c
@@ -199,11 +199,26 @@ _eo_signals_cb_added_deled(void *data EINA_UNUSED, const
Efl_Event *event)
fail_if(callback_array->func != _eo_signals_cb_added_deled);
}
-EFL_CALLBACKS_ARRAY_DEFINE(_eo_signals_callbacks,
-{ EV_A_CHANGED, _eo_signals_a_changed_cb },
-{ EV_A_CHANGED, _eo_signals_a_changed_cb2 },
-{ EV_A_CHANGED, _eo_signals_a_changed_never },
-{ EFL_EVENT_DEL, _eo_signals_efl_del_cb });
+// We don't use the EFL_CALLBACKS_ARRAY_DEFINE macro because
+// we need the callbacks be called in a specific order
+static Efl_Callback_Array_Item *
+_eo_signals_callbacks(void)
+{
+ static Efl_Callback_Array_Item items[] =
+ {
+ { EV_A_CHANGED, _eo_signals_a_changed_cb },
+ { EV_A_CHANGED, _eo_signals_a_changed_cb2 },
+ { EV_A_CHANGED, _eo_signals_a_changed_never },
+ { 0, _eo_signals_efl_del_cb },
+ { 0, 0 },
+ };
+
+ // On Windows, because _EFL_EVENT_DEL is a symbol exported
+ // from the DLL, we can't assign from a context expression
+ items[3].desc = EFL_EVENT_DEL;
+
+ return items;
+}
EFL_START_TEST(eo_signals)
{
_______________________________________________
enlightenment-devel mailing list
enlightenment-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
