Summary:
Patch from a series of patches to rename EAPI symbols to specific
library DSOs.
= The Rationale =
This patch is from a series of patches to rename EAPI symbols to
specific library DSOs.
EAPI was designed to be able to pass
`__attribute__ ((visibility ("default")))` for symbols with
GCC, which would mean that even if -fvisibility=hidden was used
when compiling the library, the needed symbols would get exported.
MSVC __almost__ works like GCC (or mingw) in which you can
declare everything as export and it will just work (slower, but
it will work). But there's a caveat: global variables will not
work the same way for MSVC, but works for mingw and GCC.
For global variables (as opposed to functions), MSVC requires
correct DSO visibility for MSVC: instead of declaring a symbol as
export for everything, you need to declare it as import when
importing from another DSO and export when defining it locally.
With current EAPI definitions, we get the following example
working in mingw and MSVC (observe it doesn't define any global
variables as exported symbols).
Example 1:
dll1:
```
EAPI void foo(void);
EAPI void bar()
{
foo();
}
```
dll2:
```
EAPI void foo()
{
printf ("foo\n");
}
```
This works fine with API defined as __declspec(dllexport) in both
cases and for gcc defining as
`__atttribute__((visibility("default")))`.
However, the following:
Example 2:
dll1:
```
EAPI extern int foo;
EAPI void foobar(void);
EAPI void bar()
{
foo = 5;
foobar();
}
```
dll2:
```
EAPI int foo = 0;
EAPI void foobar()
{
printf ("foo %d\n", foo);
}
```
This will work on mingw but will not work for MSVC. And that's why
LIBAPI is the only solution that works for MSVC.
Co-authored-by: João Paulo Taylor Ienczak Zanette <jpaulotiz@gmail.com>
Co-authored-by: Lucas Cavalcante de Sousa <lucks.sousa@gmail.com>
Co-authored-by: Ricardo Campos <ricardo.campos@expertise.dev>
Reviewers: jptiz, lucas, vtorri, woohyun
Reviewed By: jptiz, lucas, vtorri
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D12203
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
Revert "eo: redo vtable mro creation"
This reverts commit b05110609b.
Revert "eo: add a generic memory allocation tracking method"
This reverts commit 44071e3102.
Revert "eo: rework vtable allocation scheme"
This reverts commit 3bd16a46f1.
Revert "eo: do not allocate extension if deleting"
This reverts commit 64f7edc7fc.
This seems to breal vector rendering in lottie:
From: Hermet Park <hermetpark@gmail.com>
To: Enlightenment developer list <enlightenment-devel@lists.sourceforge.net>
Subject: Re: [E-devel] [EGIT] [core/efl] master 02/05: eo: rework
vtable allocation scheme
This patch occurs memory corruption, vector crashes :(
Here is a sample if you'd like to see it.
https://phab.enlightenment.org/F3858944
Summary:
with this commit a new way of allocating vtables arrived.
The old mechnism was to allocate a table big enough to carry *all*
functions at once, in order to not allocate that much memory for
functions that are not implemented on a specific klass, dichchains have
been used, which can be seens as a 2D matrix, where columns are only
allocated if min 1 entry needs to be written, this may have been a good
way to allocate back in the day when all this with eo started, however,
it showed to not pay off.
With this new way, we allocate a array of arrays. the first lvl array is
carrying enough slots, that *all* up to the time defined
interfaces/classes/abstracts/mixins can be implemented. The second lvl
array then has exactly the size of the defined APIs. The second lvl
array is obviously only allocated if needed.
When comparing the two methods, i messured two things, the usage based
on memory allocation for vtable-way-1 and vtable-way-2. Additionally, i
checked the overall memory usage of elementary_test using pmap. The
first messurement is a little bit more exact. The second messurement is
more biased, but captures the whole picture.
Memory allocation tracking:
vtable-way-1 - vtable-way-2 = 74680 Byte
Pmap memory tracking:
vtable-way1 - vtable-way-2 = 217088 Byte
The second messurement shows a bigger impact, likely because this is
also showing off all the sideeffects that we are taking place due to
fewer allocations.
Depends on D11524
Reviewers: zmike, tasn, stefan_schmidt, woohyun, cedric, raster
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D11535
This patch introduce a small hash (64 or 32bits) that cache all existing
Efl_Event_Description callback handler registered on an object. It slightly
reduce the time needed to do an unecessary call and cost just a few bytes
per object.
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D10475
As in the previous commit explained, we want to get rid of class
functions in eo, and make them just c functions right away.
This commit removes the class parameter from the eo_class_function_set
call, and adjusts the tests to not depend on class functions anymore.
Class functions are now not tested anymore, tests that used them as a
way to test *things* are adjusted to test them now with object
functions, tests that just tested the working of class functions are
dropped.
This fixes T7675.
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D7902
The next commit will bring support for something like reflection. This
commit prepares the whole tree for getting another argument in
efl_class_functions_set.
ref T7681
Differential Revision: https://phab.enlightenment.org/D7882
this enables an app or a platform to add an override for a given class,
then return a different object when that class is created. the benefit is
that a class can be internally customized by the app without needing to
modify upstream versions of that class
@feature
fix T7516
Reviewed-by: Marcel Hollerbach <marcel-hollerbach@t-online.de>
Differential Revision: https://phab.enlightenment.org/D7702
failing to call this prior to the thread invocation would result in a CRI
from class initialization in a thread
ref T7003
Differential Revision: https://phab.enlightenment.org/D6333
Summary:
build with this disabled is just not working as the thread model of eo
falls apart. The threadmodel is required to have eo_id, as it decoded
the thread information in the eo_id, which is not working without eo_id.
This also fixes the testcases that have never been executed due to the
fact of the missing HAVE_EO_ID
fix T6610
Depends on D6327
Reviewers: devilhorns, zmike
Reviewed By: zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Maniphest Tasks: T6610
Differential Revision: https://phab.enlightenment.org/D6328
individual tests should not need to explicitly call init/shutdown functions
in most cases, and many did not properly do this anyway
see followup commit which resolves some issues with eina tests
ref T6813
ref T6811
Reviewed-by: Stefan Schmidt <stefan@osg.samsung.com>
efl_check.h must be included and the EFL_START/END_TEST macros must be
used in place of normal START/END_TEST macros
timing is enabled when TIMING_ENABLED is set
https://phab.enlightenment.org/w/improve_tests/
Reviewed-by: Stefan Schmidt <stefan@osg.samsung.com>
so it was listening for cb adds and dels... and or del of any cb
except the cb add/del catcher was done.. it would fail...
but ... the test actually added other cbs than this ... like:
efl_event_callback_array_priority_add(obj, _eo_signals_callbacks(),
-100, (void *) 1);
a while array of cb's:
{ 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 });
none of which were _eo_signals_cb_added_deled. i am amazed it passed
before...
now switch its checks to check for itself and then check for anything
BUT itself...
so the MAIN loop is actually an efl.app object. which inherits from
efl.loop. the idea is that other loops in threads will not be efl.app
objects. thread on the creator side return an efl.thread object.
inside the thread, like the mainloop, there is now an efl.appthread
object that is for all non-main-loop threads.
every thread (main loop or child) when it spawns a thread is the
parent. there are i/o pipes from parnet to child and back. so parents
are generally expected to, if they want to talk to child thread, so
use the efl.io interfaces on efl.thread, and the main loop's elf.app
class allows you to talk to stdio back to the parent process like the
efl.appthread does the same using the efl.io interfaces to talk to its
parent app or appthread. it's symmetrical
no tests here - sure. i have been holding off on tests until things
settle. that's why i haven't done them yet. those will come back in a
subsequent commit
for really quick examples on using this see:
https://phab.enlightenment.org/F2983118https://phab.enlightenment.org/F2983142
they are just my test code for this.
Please see this design document:
https://phab.enlightenment.org/w/efl-loops-threads/
Triggered after (almost) complete destruction of the object.
Not called "deleted" because the other event is already "del".
I don't like "destruct" much but this follows the terminology of
"constructor" / "destructor".
@feature
This API is meant to be used by parts only, and by bindings dealing with
part objects. This patch fixes make check which got broken in the after
the previous one (cxx).
Before screaming in horror (C++...) here's why we may need this:
Efl.Part.part API returns an object that is by definition valid for a
single function call only. Enforcing this in practice is actually quite
hard as all implementation functions must manually take care of the
life-cycle. This is a lot of code in many places and a lot of
opportunities to forget to properly handle that life-cycle. Also, this
means any invalid function call on a part will leak an object.
This API absolutely must remain either "internal" or "beta" and
definitely not become abused by applications. On top of that such an API
can cause great trouble for bindings like C++. As a consequence, only
specially crafted APIs like efl_part() should return an object marked as
auto_unref.
Alternatively this API could be defined in Eo.h or some other
Eo_Internal.h. I placed it in efl_object.eo because it's much more
convenient :) (read: I'm lazy)
****
Performance notes:
Tested with clang & gcc (with -O2), I had a look at the output of perf
top, in particular the asm view. I used eo_bench in a loop. My
conclusions are:
- EINA_LIKELY/UNLIKELY actually works. The jump statement varies
according to the expectation. I highly doubt all those ugly goto in
eo.c / Eo.h are even useful.
- The impact of auto_unref on a call_resolve is so small it doesn't even
appear in the trace. It is significant inside call_end, though
(obviously, that function is just a few lines long). That function
accounts for ~1% to ~4% of all CPU time. The impact of auto_unref in
call_end is ~4% of the function time. This means ~0.16% of all CPU
time (worst measured case). _efl_object_op_api_id_get simply doesn't
show up because of caching, so the extra check there is negligible.
PS: I also tested EINA_LIKELY/UNLIKELY by compiling with -O2 and looking
at the output with objdump. The flag is well respected, and the jump
instructions are what you would expect (no jump for LIKELY and jump for
UNLIKELY). Conclusion: The goto's in eo.c only make the code harder to
read...
This makes sure the object is actually still alive and kicking before
returning any data. Otherwise the "safe" word is a bit of an abuse...
Ref T5869
@fix
This allows two things:
- adding new override functions on an object that already has
overrides
- resetting a specific function (or list of functions) to the
parent class implementation by passing NULL as implementation
Fixes T5580
@feature
This is a safe version of efl_data_scope_get, meaning that it will
return NULL if the object is not of the required type, or if there
is no data for that class, or if the given class was used as an
interface (and isn't a mixin).
@feature
This is similar to efl_super but the specified class is the one
we want to call the function on. This is similar to dynamic_cast<>
in C++.
Note: both efl_super() and efl_cast() need documentation!
This is an experimental feature.
Fixes T5311
@feature
Maniphest Tasks: T5311
Differential Revision: https://phab.enlightenment.org/D4797
This happens with shared objects.
The situation seems to be:
1. object has composited object a of class A in thread 1
2. call something on object a from thread 2, deadlock
In fact, do anything from thread 2 on a shared object and you deadlock.
Felipe Magno de Almeida
Carsten Haitzler
Wander Lairson Costa
Marcel Hollerbach
Cedric Bail
Mike Blumenkrantz
Cedric BAIL
Mike Blumenkrantz
Jean-Philippe Andre
Andy Williams
Tom Hacohen