GNU bug report logs - #9316
AIX and JNI: .so versus .a

From: "Daniel Richard G." <skunk <at> iSKUNK.ORG>
To: bug-libtool <at> gnu.org
Subject: AIX and JNI: *.so versus *.a
Date: Wed, 17 Aug 2011 17:58:26 -0400

Hello,

I am using Libtool 2.4 to build Java JNI modules on AIX.

I've noticed that on AIX, a shared library is initially built as an .so
file via "gcc -shared", but then this .so is packed up into an .a file
using ar(1). This is the case even if -export-dynamic -module is
specified.

When I go to use these JNI modules, however, the JRE makes very clear
that it doesn't like .a files:

Exception in thread "main" java.lang.UnsatisfiedLinkError: jni_test (    0509-022 Cannot load module /tmp/jni/libjni_test.a.
        0509-103   The module has an invalid magic number.)
        at java.lang.ClassLoader.loadLibraryWithPath(ClassLoader.java:994)
        at java.lang.ClassLoader.loadLibraryWithClassLoader(ClassLoader.java:958)
        at java.lang.System.loadLibrary(System.java:453)
        at com.example.jnitest.JniTest.<clinit>(JniTest.java:21)
        at java.lang.J9VMInternals.initializeImpl(Native Method)
        at java.lang.J9VMInternals.initialize(J9VMInternals.java:196)
        at JniTest.main(JniTest.java:56)

If I extract the .so files from the .a files, however, the JRE happily
uses those. Currently, I am using an install-exec-hook rule to perform
this step during "make install". There does not appear to be a Libtool
option to skip the use of ar(1), which would be a preferable approach.

--Daniel

-- 
NAME = Daniel Richard G.     _\|/_    Remember, skunks
MAIL = skunk <at> iSKUNK.ORG     (/o|o\) _- don't smell bad---
MAIL+= skunk <at> alum.MIT.EDU   < (^),>     it's the people who
WWW  = (not there yet!)      /   \      annoy us that do!

From: "Daniel Richard G." <skunk <at> iSKUNK.ORG>
To: "Peter O'Gorman" <peter <at> pogma.com>
Cc: 9316 <at> debbugs.gnu.org
Subject: Re: bug#9316: AIX and JNI: *.so versus *.a
Date: Sat, 20 Aug 2011 00:23:17 -0400

On Fri, 2011 Aug 19 22:55-0500, Peter O'Gorman wrote:
>
> If you add -Wl,-brtl to LDFLAGS at configure time, libtool will build
> plain .so's

Interesting. This bit from doc/notes.txt in the Libtool source
distribution seems relevant:

   * On AIX there are two different styles of shared linking, one in
     which symbols are bound at link-time and one in which symbols are
     bound at runtime only, similar to ELF.  In case of doubt use
     `LDFLAGS=-Wl,-brtl' for the latter style.

Why isn't Libtool itself already adding this flag when -module is
specified?

From: Peter O'Gorman <peter <at> pogma.com>
To: Daniel Richard G. <skunk <at> iSKUNK.ORG>
Cc: 9316 <at> debbugs.gnu.org
Subject: Re: bug#9316: AIX and JNI: *.so versus *.a
Date: Fri, 19 Aug 2011 22:55:45 -0500

On Aug 17, 2011, at 4:58 PM, Daniel Richard G. wrote:

> Hello,
> 
> I am using Libtool 2.4 to build Java JNI modules on AIX.

Hi,

If you add -Wl,-brtl to LDFLAGS at configure time, libtool will build plain .so's

Peter

> 
> I've noticed that on AIX, a shared library is initially built as an .so
> file via "gcc -shared", but then this .so is packed up into an .a file
> using ar(1). This is the case even if -export-dynamic -module is
> specified.
> 
> When I go to use these JNI modules, however, the JRE makes very clear
> that it doesn't like .a files:
> 
> Exception in thread "main" java.lang.UnsatisfiedLinkError: jni_test (    0509-022 Cannot load module /tmp/jni/libjni_test.a.
>        0509-103   The module has an invalid magic number.)
>        at java.lang.ClassLoader.loadLibraryWithPath(ClassLoader.java:994)
>        at java.lang.ClassLoader.loadLibraryWithClassLoader(ClassLoader.java:958)
>        at java.lang.System.loadLibrary(System.java:453)
>        at com.example.jnitest.JniTest.<clinit>(JniTest.java:21)
>        at java.lang.J9VMInternals.initializeImpl(Native Method)
>        at java.lang.J9VMInternals.initialize(J9VMInternals.java:196)
>        at JniTest.main(JniTest.java:56)
> 
> If I extract the .so files from the .a files, however, the JRE happily
> uses those. Currently, I am using an install-exec-hook rule to perform
> this step during "make install". There does not appear to be a Libtool
> option to skip the use of ar(1), which would be a preferable approach.
> 
> 
> --Daniel
> 
> 
> -- 
> NAME = Daniel Richard G.     _\|/_    Remember, skunks
> MAIL = skunk <at> iSKUNK.ORG     (/o|o\) _- don't smell bad---
> MAIL+= skunk <at> alum.MIT.EDU   < (^),>     it's the people who
> WWW  = (not there yet!)      /   \      annoy us that do!
> 
> 
> 
> _______________________________________________
> Bug-libtool mailing list
> Bug-libtool <at> gnu.org
> https://lists.gnu.org/mailman/listinfo/bug-libtool
> 

--
Peter O'Gorman
http://pogma.com

GNU bug report logs - #9316 AIX and JNI: *.so versus *.a

GNU bug report logs - #9316
AIX and JNI: .so versus .a