[Arm-dev] java-1.8.0 not building on armv5

Wed Aug 30 14:35:02 UTC 2017
Fabian Arrotin <arrfab at centos.org>

On 30/08/17 10:14, Fabian Arrotin wrote:
> On 29/08/17 19:50, Jacco Ligthart wrote:
>> On 08/29/17 10:19, Fabian Arrotin wrote:
>>> On 27/08/17 20:13, Jacco Ligthart wrote:
>>>> Hi Fabian,
>>>>
>>>> I fail to build java-1.8.0-openjdk on armv5.
>>>>
>>>> I get errors regarding gdb:
>>>>
>>>> + gdb -q
>>>> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.141-2.b16.el7.arm/openjdk/build/jdk8.build/images/j2sdk-image/bin/java
>>>> + tee gdb.out
>>>> Reading symbols from
>>>> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.141-2.b16.el7.arm/openjdk/build/jdk8.build/images/j2sdk-image/bin/java...done.
>>>> (gdb) Signal        Stop        Print   Pass to program Description
>>>> SIGSEGV       No        No      Yes             Segmentation fault
>>>> (gdb) Signal        Stop        Print   Pass to program Description
>>>> SIGILL        No        No      Yes             Illegal instruction
>>>> (gdb) (gdb) Breakpoint 1 (javaCalls.cpp:1) pending.
>>>> (gdb) >>>(gdb) Starting program:
>>>> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.141-2.b16.el7.arm/openjdk/build/jdk8.build/images/j2sdk-image/bin/java
>>>> -version
>>>> No source file named javaCalls.cpp.
>>>> [Thread debugging using libthread_db enabled]
>>>> Using host libthread_db library "/lib/libthread_db.so.1".
>>>> warning: Probes-based dynamic linker interface failed.
>>>> Reverting to original interface.
>>>> Cannot access memory at address 0xc2400000
>>>> Missing separate debuginfos, use: debuginfo-install
>>>> glibc-2.17-157.el7.4.redsleeve.armv5tel libgcc-4.8.5-11.el7.armv5tel
>>>> zlib-1.2.7-17.el7.armv5tel
>>>> (gdb) quit
>>>> A debugging session is active.
>>>>
>>>>
>>>> This is all related to a new section about gdb in the SPEC-file:
>>>>
>>>> # Make sure gdb can do a backtrace based on line numbers on libjvm.so
>>>> gdb -q "$JAVA_HOME/bin/java" <<EOF | tee gdb.out
>>>> handle SIGSEGV pass nostop noprint
>>>> handle SIGILL pass nostop noprint
>>>> set breakpoint pending on
>>>> break javaCalls.cpp:1
>>>> commands 1
>>>> backtrace
>>>> quit
>>>> end
>>>> run -version
>>>> EOF
>>>> grep 'JavaCallWrapper::JavaCallWrapper' gdb.out
>>>>
>>>>
>>>> Now my question, does this build on armv7 ?
>>>>
>>>> Jacco
>>> Hi Jacco,
>>>
>>> I'm just back from "offline/holidays" mode but I saw the issue before
>>> leaving, and have still to find the fix.
>>>
>>> What is interesting is that it seems a different error message on armv7
>>> (from
>>> https://armv7.dev.centos.org/rpmbuild/c71708-pass-1/19329-java-1.8.0-openjdk-1.8.0.131-11.b12.el7/armv7hl/build.log)
>>> :
>>>
>>> Caused by: java.io.FileNotFoundException:
>>> /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.141-2.6.10.5.el7.arm/jre/lib/resources.jar
>>> (Permission denied)
>>>
>>> Don't know why it grabbed 1.7 in the first place in the mock chroot
>>> (DEBUG util.py:417:   -->
>>> 1:java-1.7.0-openjdk-devel-1.7.0.141-2.6.10.5.el7.armv7hl) , so maybe
>>> some different mock config file is needed to enforce only fetching 1.8 ?
>>> I'll give it a try later
>> Strange that you have a different error.
>>
>> I do know why it grabs 1.7. The Spec file states:
>> # Use OpenJDK 7 where available (on RHEL) to avoid
>> # having to use the rhel-7.x-java-unsafe-candidate hack
>> %if 0%{?rhel}
>> BuildRequires: java-1.7.0-openjdk-devel
>> %else
>> BuildRequires: java-1.8.0-openjdk-devel
>> %endif
>>
>> Apparently we qualify as "rhel" here.
>>
>> Jacco
> 
> Hi Jacco,
> 
> So two things :
> 
> I tried to build it against 7.3.1611+updates and then I confirm what
> you're seeing with gdb
> (https://armv7.dev.centos.org/rpmbuild/c71611-updates-1/19786-java-1.8.0-openjdk-1.8.0.131-11.b12.el7/armv7hl/build.log)
> :
> 
> + gdb -q
> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.131-11.b12.el7.arm/openjdk/build/jdk8.build/imag
> es/j2sdk-image/bin/java
> Reading symbols from
> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.131-11.b12.el7.arm/openjdk/build/jdk
> 8.build/images/j2sdk-image/bin/java...done.
> (gdb) Signal        Stop        Print   Pass to program Description
> SIGSEGV       No        No      Yes             Segmentation fault
> (gdb) Signal        Stop        Print   Pass to program Description
> SIGILL        No        No      Yes             Illegal instruction
> (gdb) (gdb) Breakpoint 1 (javaCalls.cpp:1) pending.
> (gdb) >>>(gdb) Starting program:
> /builddir/build/BUILD/java-1.8.0-openjdk-1.8.0.131-11.b12.el7.arm/openjdk/build/jdk8.build/images/j2sdk-image/bin/java
> -version
> No source file named javaCalls.cpp.
> [Thread debugging using libthread_db enabled]
> Using host libthread_db library "/lib/libthread_db.so.1".
> Missing separate debuginfos, use: debuginfo-install
> glibc-2.17-157.el7.5.armv7hl libgcc-4.8.5-11.el7.armv7hl
> zlib-1.2.7-17.el7.armv7hl
> (gdb) warning: Probes-based dynamic linker interface failed.
> Reverting to original interface.
> Cannot access memory at address 0xc2400000
> quit
> A debugging session is active.
>         Inferior 1 [process 4802] will be killed.
> Quit anyway? (y or n) [answered Y; input not from terminal]
> + grep JavaCallWrapper::JavaCallWrapper gdb.out
> RPM build errors:
> error: Bad exit status from /var/tmp/rpm-tmp.LcfCwP (%check)
>     Bad exit status from /var/tmp/rpm-tmp.LcfCwP (%check)
> Child return code was: 1
> EXCEPTION: [Error()]
> 
> I also checked the other log
> (https://armv7.dev.centos.org/rpmbuild/c71708-pass-1/19329-java-1.8.0-openjdk-1.8.0.131-11.b12.el7/armv7hl/build.log)
> about resources.jar file missing, when it finds 1.7 and I exploded the
> rpm through rpm2cpio to verify the content and yes, there is a
> permission issue in the -headless package :
> /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.141-2.6.10.5.el7.arm/jre/lib/resources.jar
> is 600 so probably not readable by mock itself (argh)
> 
> Have now to investigate if the perm issue is due to that package build,
> or if the .spec needs some changes to force better permissions.
> 

So the issue with resources.jar is coming from an interim build that I'm
missing, so have to rebuild it first  (from the .spec)
 - Reset permissions of resources.jar to avoid it only being readable by
root (PR1437).
 - Resolves: rhbz#1350042 (but that one is private so can't see what was
changed)

Seem linked to https://bugzilla.redhat.com/show_bug.cgi?id=1207129 though

So I tried to rebuild older 1.7 pkg but it fails with "duplicate case
value" . I then found this bug report, and Jacco is included in that bug
report, the IT world is small after all :-) :
http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2942

I tried fixing 1.7 openjdk with the
+  ARM32JIT="false" \
patch for the make to ensure that it doesn't build arm32 JIT, but for
strange reason it still tries to build it and fails again ... some other
investigations required to rebuild older 1.7 from pre-7.4, then the one
from 7.4 (with the resources.jar permission issue gone) and then 1.8
(hopefully) .. of course it happens on java, and on armhp, where each
build needs *several* hours ....


-- 
Fabian Arrotin
The CentOS Project | http://www.centos.org
gpg key: 56BEC54E | twitter: @arrfab

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: OpenPGP digital signature
URL: <http://lists.centos.org/pipermail/arm-dev/attachments/20170830/907a65b3/attachment-0006.sig>