UnsatisfiedLinkError for linux-arm when using jna 4.2.0 from Maven repo

[antonio-dibacco]

I get the following exception with jna 4.2.0 for linux arm platform 32 bit:

Exception in thread "main" java.lang.UnsatisfiedLinkError: /tmp/jna-3506402/jna8529987963707458054.tmp: /tmp/jna-3506402/jna8529987963707458054.tmp: internal error
at java.lang.ClassLoader$NativeLibrary.load(Native Method)
at java.lang.ClassLoader.loadLibrary0(Unknown Source)
at java.lang.ClassLoader.loadLibrary(Unknown Source)
at java.lang.Runtime.load0(Unknown Source)
at java.lang.System.load(Unknown Source)
at com.sun.jna.Native.loadNativeDispatchLibraryFromClasspath(Native.java:851)
at com.sun.jna.Native.loadNativeDispatchLibrary(Native.java:826)
at com.sun.jna.Native.(Native.java:140)
at com.sun.jna.Pointer.(Pointer.java:41)
at com.sun.jna.Structure.(Structure.java:2078)
at Net.main(Net.java:32)

[Zubnix]

Can confirm. Same issue here.

[twall]

Are you running v5 or v7? hf or soft float?

I’m pretty sure the JNA library for linux/arm is compiled for v7+hf (android defaults to v5+soft float)

On Dec 11, 2015, at 6:07 PM, Zubnix [email protected] wrote:

Can confirm. Same issue here.

—
Reply to this email directly or view it on GitHub.

[Zubnix]

I've tested on rpi 1 model B using latest raspbian, which is I believe is armv6hf. I've also seen the issue on a freescale imx6, which I believe was running an armv7 hf yocto build.

[CmptrExpr]

I'm seeing the same problem under Windows 2012 with JNA 4.2.1 and 4.2.2.

[twall]

You’re seeing UnsatisfiedLinkError or “internal error”?

On Apr 7, 2016, at 11:42 AM, Steve [email protected] wrote:

I'm seeing the same problem under Windows 2012 with JNA 4.2.1 and 4.2.2.

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub

[CmptrExpr]

UnsatisfiedLinkError

Steve Freeman

On Thu, Apr 7, 2016 at 4:03 PM, Timothy Wall [email protected]
wrote:

You’re seeing UnsatisfiedLinkError or “internal error”?

On Apr 7, 2016, at 11:42 AM, Steve [email protected] wrote:

I'm seeing the same problem under Windows 2012 with JNA 4.2.1 and 4.2.2.

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#538 (comment)

[twall]

Not the same error. arm/internal error != windows unsatisfied link error.

Please post to jna-users first (with full context and stack trace) and open a new issue here only if it’s determined to be a bug and not a configuration issue.

On Apr 7, 2016, at 4:23 PM, Steve [email protected] wrote:

UnsatisfiedLinkError

Steve Freeman

On Thu, Apr 7, 2016 at 4:03 PM, Timothy Wall [email protected]
wrote:

You’re seeing UnsatisfiedLinkError or “internal error”?

On Apr 7, 2016, at 11:42 AM, Steve [email protected] wrote:

I'm seeing the same problem under Windows 2012 with JNA 4.2.1 and 4.2.2.

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#538 (comment)

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub

[matthiasblaesing]

Please test with 4.2.2 - if it fails, please:

• make sure only one copy of JNA is on the class path
• report the exact version of the JDK/JRE (including bitness) and
• exact processor + linux distribution you are running

I'm closing this as not enough information is provided. Feel free to reopen with further information.

[bondolo]

Seeing this problem with JNA 4.2.2 with JDK 8u6 (32bit, softfloat) on ARM7l softfloat (32-bit). The OS is TimeSys Linux 3.0 (softfloat).

[matthiasblaesing]

yes - I'm sorry the ARM builds are hardfloat. If you can give be a pointer how to determine hardfloat/softfloat at runtime, I'm willing to do softfloat builds (in fact I already did by accident).

[bondolo]

Sorry, I don't know the answer to that. Earlier JNA builds were softfloat compatible, 4.0.0, which is what I am upgrading from did include ARM7 softfloat support.

[bondolo]

This is what I was able to find regarding determining armel vs armfp:

https://www.raspberrypi.org/forums/viewtopic.php?f=33&t=20873

[bondolo]

And this which appears to be the best solution:

https://bugs.openjdk.java.net/browse/JDK-8005545

gnueabi GNU softfp EABI
gnueabihf GNU hard float EABI
androideabi Android EABI

[bondolo]

For java 7 on ARM I don't believe there was ever an official hardfloat build so if the property doesn't exist assume softfloat

[matthiasblaesing]

Thank you - this looks also matching:

https://blogs.oracle.com/jtc/entry/is_it_armhf_or_armel

With regard to past builds, I just fetched from maven central: 4.0.0, 4.2.0, 4.2.2, 4.3.0. They are all hardfloat:

matthias@athena:~/x$ readelf -A x-jna-4.3.0.jar/com/sun/jna/linux-arm/libjnidispatch.so  | grep VFP
  Tag_FP_arch: VFPv2
  Tag_ABI_VFP_args: VFP registers
matthias@athena:~/x$ readelf -A x-jna-4.2.2.jar/com/sun/jna/linux-arm/libjnidispatch.so  | grep VFP
  Tag_FP_arch: VFPv2
  Tag_ABI_VFP_args: VFP registers
matthias@athena:~/x$ readelf -A x-jna-4.2.0.jar/com/sun/jna/linux-arm/libjnidispatch.so  | grep VFP
  Tag_FP_arch: VFPv2
  Tag_ABI_VFP_args: VFP registers
matthias@athena:~/x$ readelf -A x-jna-4.0.0.jar/com/sun/jna/linux-arm/libjnidispatch.so  | grep VFP
  Tag_FP_arch: VFPv2
  Tag_ABI_VFP_args: VFP registers

[bondolo]

I just re-verified that replacing the 4.2.2 dependency with 4.0.0 works correctly for purejavacomm. I can't explain the difference in behaviour between 4.0.0. and 4.2.2

diff -uwd <(readelf -A 4.0.0.libjnidispatch.so) <(readelf -A 4.2.2.libjnidispatch.so)

shows that they were built with the same options. ie. no difference.

The repo version of libjnidispatch is built differently but I haven't actually tested that:

diff -uwd <(readelf -A 4.2.2.libjnidispatch.so) <(readelf -A libjnidispatch.so)
--- /dev/fd/63 2017-01-16 21:24:27.897202513 +0000
+++ /dev/fd/62 2017-01-16 21:24:27.905015003 +0000
@@ -13,5 +13,4 @@
Tag_ABI_enum_size: int
Tag_ABI_HardFP_use: SP and DP
Tag_ABI_VFP_args: VFP registers

• Tag_ABI_optimization_goals: Aggressive Speed
• Tag_DIV_use: Not allowed

• Tag_CPU_unaligned_access: v6

[matthiasblaesing]

Where does the other libjnidispatch come from? If that was build for your system, there is no bug... Please move the discussion to the jna-user mailinglist. A hijacked bug is not good for a discussiin. Am 16. Januar 2017 22:28:27 MEZ schrieb Mike Duigou <[email protected]>:

…

I just re-verified that replacing the 4.2.2 dependency with 4.0.0 works correctly for purejavacomm. I can't explain the difference in behaviour between 4.0.0. and 4.2.2 diff -uwd <(readelf -A 4.0.0.libjnidispatch.so) <(readelf -A 4.2.2.libjnidispatch.so) shows that they were built with the same options. ie. no difference. The repo version of libjnidispatch _is_ built differently but I haven't actually tested that: diff -uwd <(readelf -A 4.2.2.libjnidispatch.so) <(readelf -A libjnidispatch.so) --- /dev/fd/63 2017-01-16 21:24:27.897202513 +0000 +++ /dev/fd/62 2017-01-16 21:24:27.905015003 +0000 @@ -13,5 +13,4 @@ Tag_ABI_enum_size: int Tag_ABI_HardFP_use: SP and DP Tag_ABI_VFP_args: VFP registers - Tag_ABI_optimization_goals: Aggressive Speed - Tag_DIV_use: Not allowed + Tag_CPU_unaligned_access: v6 -- You are receiving this because you modified the open/close state. Reply to this email directly or view it on GitHub: #538 (comment)

-- Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

[bondolo]

The other libjnidispatch is 4.0.0. I will resume on jna-user mailinglist