Bugzilla Bug 87774

During emerge this error comes up and compilation fails:

make[3]: Leaving directory `/var/tmp/portage/pilot-link-0.11.8-r1/work/pilot-link-0.11.8/bindings/Java'
cp libjpisock.so /opt/sun-jdk-1.5.0.01/jre/lib/i386/
ACCESS DENIED  open_wr:   /opt/sun-jdk-1.5.0.01/jre/lib/i386/libjpisock.so
cp: regular file ,,/opt/sun-jdk-1.5.0.01/jre/lib/i386/libjpisock.so" cannot be created: no access rights

Of course this doesn't work. I don't have /opt/sun-jdk-1.5.0.01/jre/lib/i386/
but I have /opt/sun-jdk-1.5.0.01/jre/lib/amd64/
so somewhere the wrong (hardcoded?) path is specified.

You could create a symlink to side-step the issue for now. Silly bug that needs
fixing upstream I guess.

OK, will create a symlink. I send a mail to the pilot-link devl list. Hopefully
they fix it. Should this be left open until then, or should it be closed?

well i've got the same issue and have created a symlink from i386 to amd64 with
no change in behaviour.  did i miss something?

its also a bad idea to install it directly into JAVA_HOME, since that dir gets
changed when there is an update to the jdk/jre

This is still an open issue, apparently.  I did some tests again today and it's
still looking for the i386 directory.

Any thoughts for a fix?  I haven't had time to build a patch for this myself.

Creating symlink seems to work for me.  Ugly though.  Wish I could find where
upstream this was hard coded as i386.

FYI:  Creating a symlink does NOT work for me.  I still get sandbox violations:

------------
cp libjpisock.so /opt/blackdown-jdk-1.4.2.01/jre/lib/i386/
ACCESS DENIED  open_wr:   /opt/blackdown-jdk-1.4.2.01/jre/lib/i386/libjpisock.so
cp: cannot create regular file `/opt/blackdown-jdk-1.4.2.01/jre/lib/i386/libjpisock.so': Permission denied
------------

*** Bug 89823 has been marked as a duplicate of this bug. ***

*** Bug 91668 has been marked as a duplicate of this bug. ***

*** Bug 92834 has been marked as a duplicate of this bug. ***

*** Bug 97094 has been marked as a duplicate of this bug. ***

There seem to be a couple of problems here...
1) the two amd64-patches patch Makefile.am But automake is never run by the
installation scripts, right? So they should patch Makefile.in to have any effect.

2) java_install_amd64.patch tries to copy to ${JAVABASE}/jre/lib/amd64/ which is
not relative to $(DESTDIR), hence the sandbox-violation, i think. Why not simply
install it in $(DESTDIR)$(libdir)/ ?

3) I don't see how the fPIC patch should take effect. It patches Makefile.in but
is applied *after* econf has been called. 

Created an attachment (id=62318) [details]
experimental ebuild to make it compile on amd64

beware, this is just a quick hack to get in to compile on my system (amd64)
- get rid of compile_amd64 patch
- new ebuild moves fPIC patch to the beginning of src_unpack()
- fPIC patch also adds $(CXXFLAGS)
- install_amd64 patch installs into $(DESTDIR)$(libdir)/

Good show!  I've been very busy lately or I would have tested this sooner.

I used your attached ebuild and patches and created a small Portage overlay for
this.  It worked like a charm.  Thanks very much for taking the time to dig into
the build files and figure out what was wrong (I hadn't had the time).

** To the Gentoo Devs:  I recommend incorporating these changes into the Portage
tree ASAP (after sufficient quality review, of course -- but the changes seem
minor enough).

Also, for anyone else that wants a copy of my Portage overlay to fix this just
let me know.

i still don't know whether adding -fPIC to CXXFLAGS is the right way to go? can
amd64 folks confirm this?

(In reply to comment #15)
> i still don't know whether adding -fPIC to CXXFLAGS is the right way to go?
can amd64 folks confirm this?

But that's not what the patch does. -fPIC is only added to the compile options
for libjpisock.c, which should be ok.

-fPIC must be applied to (and only to) shared objects, generally adding -fPIC
to
C(XX)FLAGS is a very bad idea. This is policy on all arches, so please apply
the
changes unconditionally.

(hint: http://dev.gentoo.org/~plasmaroo/devmanual//archs/amd64/ )

great. this should finally fix it. thanks for the patch and also the amd64 devs
for pointing me to some 
docs about fPIC.  i'm committing this into -r1 rather than a revision bump
since it doesn't add any new 
features, just fixes compile problems.

*** Bug 111434 has been marked as a duplicate of this bug. ***