Example app hangs on startup

[alexeyknyshev]

Gdb backtrace of hanged thread. Looks like hang at libkiui.so dynamic lib start:

(gdb) bt

0 syscall () at ../sysdeps/unix/sysv/linux/x86_64/syscall.S:38

1 0x00007ffff63e3023 in __cxa_guard_acquire () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6

2 0x00007ffff79c6132 in mk::Type::cls() () from /usr/local/lib/libkiui.so

3 0x00007ffff79c5ebc in mk::Type::Type() () from /usr/local/lib/libkiui.so

4 0x00007ffff79c6150 in mk::Type::cls() () from /usr/local/lib/libkiui.so

5 0x00007ffff79c5ebc in mk::Type::Type() () from /usr/local/lib/libkiui.so

6 0x00007ffff79cf05d in mk::Type& mk::typecls() () from /usr/local/lib/libkiui.so

7 0x00007ffff79d21b8 in mk::Dispatch<mk::ValueWidget, bool&, &(mk::WValue& mk::valueWidget<bool&, mk::Input >(mk::Sheet*, mk::Lref&, bool&))>::Dispatch() () from /usr/local/lib/libkiui.so

8 0x00007ffff79cee44 in __static_initialization_and_destruction_0 () from /usr/local/lib/libkiui.so

9 0x00007ffff79cefb1 in _GLOBAL__sub_I_mkInput.cpp () from /usr/local/lib/libkiui.so

10 0x00007ffff7dea9fa in call_init (l=, argc=argc@entry=1, argv=argv@entry=0x7fffffffe0c8,

env=env@entry=0x7fffffffe0d8) at dl-init.c:78

11 0x00007ffff7deaae3 in call_init (env=0x7fffffffe0d8, argv=0x7fffffffe0c8, argc=1, l=) at dl-init.c:36

12 _dl_init (main_map=0x7ffff7ffe1a8, argc=1, argv=0x7fffffffe0c8, env=0x7fffffffe0d8) at dl-init.c:126

13 0x00007ffff7ddd1ca in _dl_start_user () from /lib64/ld-linux-x86-64.so.2

14 0x0000000000000001 in ?? ()

15 0x00007fffffffe3dc in ?? ()

16 0x0000000000000000 in ?? ()

[alexeyknyshev]

I try to figure out root of problem 426db21 hangs

[hugoam]

Does that fix it ? f7d10df (It's the path set by CMake that might have been wrong, if you don't want to rerun cmake you can modify KIUI_EXAMPLE_RESSOURCE_PATH manually and tell me if that fixes it)

The backtrace is really weird though. I don't see what could go wrong so early...

[hugoam]

Oooh wait I think I'm starting to seeing the problem in Type(), calling Type::cls() from Type::cls()

[alexeyknyshev]

I'll be able to check at the morning. Now I'm lying in bed :р

[hugoam]

No problem. I think I fixed it in 16b7c4d, keep me posted.

[alexeyknyshev]

Looks like it's not connected with Type() bug. Problem has not gone. I'm trying to figure out...

[hugoam]

If the application hangs, it should be possible to pause the debugger and see where exactly the hang happens. Could you try to give me this information so that I try to understand what is going on ?

[hugoam]

Nevermind. The data path was not set properly between win32 and linux 4ef1ec4 and af54385. Tell me if that solves the problem.

[alexeyknyshev]

Hang callstack, hangs in static initialization of Type ty(0); static Type& cls() { static Type ty(0); return ty; }

0 syscall /usr/lib/debug/lib/x86_64-linux-gnu/libc-2.19.so 38 0x7ffff5b91fd9
1 cxa_guard_acquire 0x7ffff63cf023
2 mk::Type::cls mkType.h 54 0x7ffff79bd748
3 mk::Indexed<mk::Type, void>::indexer mkIndexer.h 78 0x7ffff79be37f
4 mk::Type::Type mkType.cpp 17 0x7ffff79bd3d4
5 mk::Type::cls mkType.h 54 0x7ffff79bd76b
6 mk::Type::Type mkType.cpp 21 0x7ffff79bd43c
7 mk::typecls mkTyped.h 42 0x7ffff79c66b6
8 mk::Dispatch<mk::ValueWidget, bool&, &mk::valueWidget<bool&, mk::Input>>::Dispatch mkDispatch.h 89 0x7ffff79c9b0a
9 static_initialization_and_destruction_0 mkDispatch.h 102 0x7ffff79c645e
10 _GLOBAL__sub_I_mkInput.cpp(void) mkInput.cpp 110 0x7ffff79c65cb
11 call_init /usr/lib/debug/lib/x86_64-linux-gnu/ld-2.19.so 78 0x7ffff7dea9fa
12 call_init /usr/lib/debug/lib/x86_64-linux-gnu/ld-2.19.so 36 0x7ffff7deaae3
13 _dl_init /usr/lib/debug/lib/x86_64-linux-gnu/ld-2.19.so 126 0x7ffff7deaae3
14 _dl_start_user 0x7ffff7ddd1ca
15 ?? 0x1 16 ?? 0x7fffffffea31
17 ??

Looks like dead look due to double static initailization. First time it acquires look and goes deeper, calls Type::cls() again and again makes static initialization, which already locked ==> deadlock as the result.

[hugoam]

Yes now I see it. I already removed one such recursion in Type() and now I see there's another one when it calls indexer(). Gonna fix it right away.

[hugoam]

Here dd5d3c8

[alexeyknyshev]

Ought to be fixed! Great! Thx a lot!