Bug 1065 - NEWT crash at WindowDriver.initIds0(): TISGetInputSourceProperty(keyboard, kTISPropertyUnicodeKeyLayoutData) returns NULL
Summary: NEWT crash at WindowDriver.initIds0(): TISGetInputSourceProperty(keyboard, kT...
Status: RESOLVED FIXED
Alias: None
Product: Newt
Classification: JogAmp
Component: macosx (show other bugs)
Version: 1
Hardware: pc_x86_64 macosx
: --- enhancement
Assignee: Sven Gothel
URL:
Depends on:
Blocks:
 
Reported: 2014-09-08 07:04 CEST by David Farler
Modified: 2014-09-08 08:31 CEST (History)
0 users

See Also:
Type: ---
SCM Refs:
c0c722b9f479412f27973ba0c4cd4a166dcb00be
Workaround: ---


Attachments
test.log (4.25 KB, text/plain)
2014-09-08 07:04 CEST, David Farler
Details
test_dbg.log.gz (118.15 KB, application/x-gzip)
2014-09-08 07:05 CEST, David Farler
Details
JRE Crash Log (40.33 KB, text/plain)
2014-09-08 07:06 CEST, David Farler
Details
OS X Crash Log (75.53 KB, text/plain)
2014-09-08 07:07 CEST, David Farler
Details
ef7ecf76e28 Debug OS X Crash Log (77.86 KB, application/octet-stream)
2014-09-08 07:50 CEST, David Farler
Details
NEWT patch .. 01 (1.25 KB, text/plain)
2014-09-08 08:07 CEST, Sven Gothel
Details

Note You need to log in before you can comment on or make changes to this bug.
Description David Farler 2014-09-08 07:04:27 CEST
When running the runtime tests or the following code, libnewt crashes when trying to create a display.

public static void main(String[] args) {
    Display display = NewtFactory.createDisplay("Five Display");
}

Test logs and crash logs attached.

System information:

MacBookPro11,3
Intel Core i7 2.3 Ghz
Intel Iris Pro 0x0d26 0x0008
NVIDIA GeForce GT 750M
LED Cinema Display (non Thunderbolt) - with laptop closed.


ProductName:	Mac OS X
ProductVersion:	10.9.4
BuildVersion:	13E28

java version "1.7.0_67"
Java(TM) SE Runtime Environment (build 1.7.0_67-b01)
Java HotSpot(TM) 64-Bit Server VM (build 24.65-b04, mixed mode)
Comment 1 David Farler 2014-09-08 07:04:43 CEST
Created attachment 634 [details]
test.log
Comment 2 David Farler 2014-09-08 07:05:53 CEST
Created attachment 635 [details]
test_dbg.log.gz
Comment 3 David Farler 2014-09-08 07:06:14 CEST
Created attachment 636 [details]
JRE Crash Log
Comment 4 David Farler 2014-09-08 07:07:03 CEST
Created attachment 637 [details]
OS X Crash Log
Comment 5 Sven Gothel 2014-09-08 07:19:37 CEST
Comment on attachment 635 [details]
test_dbg.log.gz

this file is double gzipped :)
Comment 6 Sven Gothel 2014-09-08 07:23:00 CEST
So we reach the native method
  Java_jogamp_newt_driver_macosx_WindowDriver_initIDs0

and crash right there .. even though it doesn't do very much.

Some JNI FindClass and getMEthod .. will throw an Exception if this fails.

JOGL seems to be valid, 2.2.0 build.

Q: Have you launched your tests from our build ?

Q: Do you have some 'anti virus' stuff running ?

You seem to use the native jar file, loaded automatically .. 
No issue here .. it seems:

+++
JarUtil: EXTRACT[5]: [newt -> ] libnewt.jnilib -> /var/folders/p6/62twq4cd43379xz_328wbkfh0000gn/T/jogamp_0000/file_cache/jln5587047091668471466/jln1056818139617755950/libnewt.jnilib: 89076 bytes, addedAsNativeLib: true

NativeLibrary.findLibrary(<newt>) (TempJarCache): /var/folders/p6/62twq4cd43379xz_328wbkfh0000gn/T/jogamp_0000/file_cache/jln5587047091668471466/jln1056818139617755950/libnewt.jnilib
JNILibLoaderBase: loadLibraryInternal(newt), TempJarCache: /var/folders/p6/62twq4cd43379xz_328wbkfh0000gn/T/jogamp_0000/file_cache/jln5587047091668471466/jln1056818139617755950/libnewt.jnilib
JNILibLoaderBase: System.load(/var/folders/p6/62twq4cd43379xz_328wbkfh0000gn/T/jogamp_0000/file_cache/jln5587047091668471466/jln1056818139617755950/libnewt.jnilib) - mode 2
+++

Dunno ..

Can you also try running an applet via Safari or Firefox ?
(no logs necessary .. just checking your launch setup)
Comment 7 Sven Gothel 2014-09-08 07:25:00 CEST
(In reply to comment #6)
> So we reach the native method
>   Java_jogamp_newt_driver_macosx_WindowDriver_initIDs0
> 
> and crash right there .. even though it doesn't do very much.
> 
> Some JNI FindClass and getMEthod .. will throw an Exception if this fails.

http://jogamp.org/git/?p=jogl.git;a=blob;f=src/newt/native/MacWindow.m;h=2bd11da3c5e4d1e5b1610f7dc2d6a8428dbb285f;hb=HEAD#l708
Comment 8 David Farler 2014-09-08 07:26:13 CEST
> Q: Have you launched your tests from our build ?

Yes, this is from 2.2.0 download (currently from https://jogamp.org/deployment/jogamp-current/archive)

> 
> Q: Do you have some 'anti virus' stuff running ?

Nope.
Comment 9 Sven Gothel 2014-09-08 07:42:21 CEST
jogamp03:~ jogamp$ xcrun  --show-sdk-path
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk
jogamp03:~ jogamp$ xcrun  --show-sdk-version
10.9
jogamp03:~ jogamp$ xcrun  --show-sdk-platform-path
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform
jogamp03:~ jogamp$ xcrun  --show-sdk-platform-version
1.1

Build node is 10.9.4 as well, Intel CPU 2.4 GHZ Core 2 Duo, 
Mac-Mini,
employs an NV card.

I also tested on a newer Macbook (Intel/NV GPU) .. no issues.
Comment 10 David Farler 2014-09-08 07:50:55 CEST
Created attachment 638 [details]
ef7ecf76e28 Debug OS X Crash Log
Comment 11 David Farler 2014-09-08 07:51:18 CEST
Xcode 5.1.1
Build version 5B1008

Apple LLVM version 5.1 (clang-503.0.40) (based on LLVM 3.4svn)
Target: x86_64-apple-darwin13.3.0
Thread model: posix
Comment 12 Sven Gothel 2014-09-08 08:07:43 CEST
Created attachment 639 [details]
NEWT patch .. 01
Comment 13 Sven Gothel 2014-09-08 08:26:08 CEST
TISGetInputSourceProperty(keyboard, kTISPropertyUnicodeKeyLayoutData) returns NULL
with keyboard layouts that have sublayout categories.
Comment 14 Sven Gothel 2014-09-08 08:27:40 CEST
(In reply to comment #13)
> TISGetInputSourceProperty(keyboard, kTISPropertyUnicodeKeyLayoutData)
> returns NULL
> with keyboard layouts that have sublayout categories.

Yes, you can simulate it by going to System Preferences -> Keyboard -> Input Sources and adding Kotoeri - that will give you a system ui menu item in top right
Comment 15 Sven Gothel 2014-09-08 08:28:02 CEST
c0c722b9f479412f27973ba0c4cd4a166dcb00be

Handle NULL result of TISGetInputSourceProperty(keyboard, kTISPropertyUnicodeKeyLayoutData), avoiding crash
Comment 16 Sven Gothel 2014-09-08 08:30:40 CEST
Closing this bug now for release 2.2.1

We may open a dependent new bug to deal with such keyboard layouts,
since this configuration will remove NEWTs functionality of 
mapping key-codes to US-ASCII normalized ones (layout independent).
Comment 17 Sven Gothel 2014-09-08 08:31:42 CEST
btw .. thank you 'magaio' for finding this bug, identifying it 
and allow us to swiftly fix at least the crash!