Simon,

On 19 October 2017 at 17:49, Simon Urbanek wrote:
| Dirk,
| 
| to clarify - this is about *R* configuration, not about rJava at all. rJava 
has nothing to do with those settings - it just uses what R was configured 
with. The only part rJava plays here is that it needs those R settings to be 
valid - just like other R packages that use Java.
| 
| As you reported the R configuration is broken, because the JAVA entry points 
to a non-existent file. I don't know why or how that happens. At least that's 
what I saw in the log:
| 
| checking Java support in R... present:
| interpreter : '/usr/lib/jvm/default-java/jre/bin/java'
| [...]
| checking whether Java run-time works... ./configure: line 3736: 
/usr/lib/jvm/default-java/jre/bin/java: No such file or directory
| no
| configure: error: Java interpreter '/usr/lib/jvm/default-java/jre/bin/java' 
does not work
| 
| I presume that setting should have been /usr/lib/jvm/default-java/bin/java -- 
or JRE is required and missing (again, I don't know what package is providing 
that /usr/lib/jvm/default-java/jre/bin/java file used in the configuration).

I am confused.   Once I added the call to 'R CMD javareconf' to be done
before 'R CMD INSTALL ...your tarball...' happens, it _builds_ but then failed
on a subsequent command that is not from me.  I could patch that.

Allow me to copy and paste again:

-----------------------------------------------------------------------------
make[1]: Entering directory '/build/rjava-0.9-9'
CDBS WARNING:    simple-patchsys.mk is deprecated since 0.4.85 - please use 
source format 3.0 (quilt) instead
set -e;   mv ./jri/tools/config.guess ./jri/tools/config.guess.cdbs-orig; cp 
--remove-destination /usr/share/misc/config.guess ./jri/tools/config.guess;
set -e;   mv ./jri/tools/config.sub ./jri/tools/config.sub.cdbs-orig; cp 
--remove-destination /usr/share/misc/config.sub ./jri/tools/config.sub;
make[1]: Leaving directory '/build/rjava-0.9-9'
set -e;   
set -e;   
R CMD javareconf
*** JAVA_HOME is not a valid path, ignoring
Java interpreter : /usr/bin/java
Java version     : 9-Debian
Java home path   : /usr/lib/jvm/java-9-openjdk-amd64
Java compiler    : /usr/bin/javac
Java headers gen.: /usr/bin/javah
Java archive tool: /usr/bin/jar

trying to compile and link a JNI program 
detected JNI cpp flags    : -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
detected JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
make[1]: Entering directory '/tmp/Rjavareconf.e0Jzrq'
gcc -std=gnu99 -I/usr/share/R/include -DNDEBUG 
-I/usr/lib/jvm/java-9-openjdk-amd64/include 
-I/usr/lib/jvm/java-9-openjdk-amd64/include/linux     -fpic  -g -O2 
-fdebug-prefix-map=/build/r-base-3.4.2=. -fstack-protector-strong -Wformat 
-Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -g  -c conftest.c -o 
conftest.o
gcc -std=gnu99 -shared -L/usr/lib/R/lib -Wl,-z,relro -o conftest.so conftest.o 
-L/usr/lib/jvm/java-9-openjdk-amd64/lib/server -ljvm -L/usr/lib/R/lib -lR
make[1]: Leaving directory '/tmp/Rjavareconf.e0Jzrq'


JAVA_HOME        : /usr/lib/jvm/java-9-openjdk-amd64
Java library path: $(JAVA_HOME)/lib/server
JNI cpp flags    : -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
Updating Java configuration in /usr/lib/R
/usr/lib/R/bin/javareconf: 406: /usr/lib/R/bin/javareconf: cannot create 
/usr/lib/R/etc/Makeconf.new: Permission denied
*** cannot create /usr/lib/R/etc/Makeconf.new
*** Please run as root if required.

debian/rules:9: recipe for target 'build/r-cran-rjava' failed
make: *** [build/r-cran-rjava] Error 1
dpkg-buildpackage: error: debian/rules build gave error exit status 2
-----------------------------------------------------------------------------

So conftest.so is built.  Can we skip the subsequent javareconf?  If so,
where?  When I glance at your configure.ac I see several ...  And the
Makefile structure is not exactly trivial or standard either.

Dirk


| Cheers,
| Simon
| 
| 
| > On Oct 19, 2017, at 4:56 PM, Dirk Eddelbuettel <e...@debian.org> wrote:
| > 
| > 
| > On 19 October 2017 at 16:31, Simon Urbanek wrote:
| > | 
| > | > On Oct 19, 2017, at 4:12 PM, Dirk Eddelbuettel <e...@debian.org> wrote:
| > | > 
| > | > 
| > | > On 19 October 2017 at 15:47, Simon Urbanek wrote:
| > | > | R CMD javareconf is always needed to register Java support with R (on 
all unix platforms).
| > | > 
| > | > s/always/sometimes/
| > | > 
| > | > The values get encoded when I build R binaries.  For many years now, 
that was
| > | > good enough. 
| > | > 
| > | > | Like I said, I don't know what deb packages do here -
| > | > 
| > | > 'R CMD INSTALL sourceTarball_1.2-3.tar.gz  -- what else do you think I 
do?
| > | > 
| > | > | I presume you must do something in your R binary and all I can 
presume is that whatever the setup is then doesn't match what's there now.
| > | > 
| > | > 500+ CRAN packages packaged _exactly_ this.  rJava is a little 
different (for
| > | > an understandable reason; RInside is different too).
| > | > 
| > | > So ... tried that, still no luck:
| > | > 
| > | > ---
| > | > [...]
| > | > R CMD javareconf
| > | > *** JAVA_HOME is not a valid path, ignoring
| > | > Java interpreter : /usr/bin/java
| > | > Java version     : 9-Debian
| > | > Java home path   : /usr/lib/jvm/java-9-openjdk-amd64
| > | > Java compiler    : /usr/bin/javac
| > | > Java headers gen.: /usr/bin/javah
| > | > Java archive tool: /usr/bin/jar
| > | > 
| > | > trying to compile and link a JNI program
| > | > detected JNI cpp flags    : -I$(JAVA_HOME)/include 
-I$(JAVA_HOME)/include/linux
| > | > detected JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
| > | > make[1]: Entering directory '/tmp/Rjavareconf.e0Jzrq'
| > | > gcc -std=gnu99 -I/usr/share/R/include -DNDEBUG 
-I/usr/lib/jvm/java-9-openjdk-amd64/include 
-I/usr/lib/jvm/java-9-openjdk-amd64/include/linux     -fpic  -g -O2 
-fdebug-prefix-map=/build/r-base-3.4.2=. -fstack-protector-strong -Wformat 
-Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -g  -c conftest.c -o 
conftest.o
| > | > gcc -std=gnu99 -shared -L/usr/lib/R/lib -Wl,-z,relro -o conftest.so 
conftest.o -L/usr/lib/jvm/java-9-openjdk-amd64/lib/server -ljvm 
-L/usr/lib/R/lib -lR
| > | > make[1]: Leaving directory '/tmp/Rjavareconf.e0Jzrq'
| > | > 
| > | > 
| > | > JAVA_HOME        : /usr/lib/jvm/java-9-openjdk-amd64
| > | > Java library path: $(JAVA_HOME)/lib/server
| > | > JNI cpp flags    : -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/linux
| > | > JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
| > | > Updating Java configuration in /usr/lib/R
| > | > /usr/lib/R/bin/javareconf: 406: /usr/lib/R/bin/javareconf: cannot 
create /usr/lib/R/etc/Makeconf.new: Permission denied
| > | > *** cannot create /usr/lib/R/etc/Makeconf.new
| > | > *** Please run as root if required.
| > | > [...]
| > | > ---
| > | > 
| > | > All this is running as root.  But /usr may be mounted read-only.
| > | > 
| > | 
| > | Right, you have to have write-access is /usr/lib/R/etc/ so it can update 
the configuration to record your Java settings.
| > 
| > Well, in general we do _not_ have that and _cannot_ assume it, just like 
CRAN
| > packages do _not_ get to write to $RHOME/bin.
| > 
| > _For Debian systems_ I offer (us, mostly) /etc/R/ with softlinks.
| > 
| > But if your package assumes /usr/lib for write only, we may be done here and
| > it will get removed.
| > 
| > | I was asking because the above it obviously non-portable - i.e. if I
| > | install Oracle Java the path will be different and also it will be 
different
| > | for Java 1.8 etc. Hence you'd need to re-run javareconf any time the Java
| > | configuration changes (thanks to Sun/Oracle for that wonderful design) -
| > | unless you have canonical place you can use. From your original settings 
it
| > | looked like you have some default-java symlink which would help here -
| > 
| > Not mine. That is the Debian Java folks.
| > 
| > | except that it has the broken path to java which is where the trouble 
started in the first place. I suspect that fixing your default config to have a 
valid path to java should be enough as long as that default setting is 
maintained at the system level.
| > 
| > No trouble _once I run R CMD javareconf_ earlier it built.  But we died in 
in
| > your post-processing, it seems.
| > 
| > Dirk
| > 
| > | Cheers,
| > | Simon
| > | 
| > | 
| > | 
| > | > Dirk
| > | > | 
| > | > | Cheers,
| > | > | Simon
| > | > | 
| > | > | 
| > | > | > On Oct 18, 2017, at 9:50 PM, Dirk Eddelbuettel <e...@debian.org> 
wrote:
| > | > | > 
| > | > | > 
| > | > | > Simon,
| > | > | > 
| > | > | > On 18 October 2017 at 21:34, Simon Urbanek wrote:
| > | > | > | Dirk,
| > | > | > | 
| > | > | > | I don't have the details but this looks like as incorrect Java 
configuration in R - the Java home is obviously incomplete as it's missing 
bin/java and/or that path is wrong. Also I'm surprised to the the odd paths 
"/usr/lib/jvm/default-java/" since at least in my installations I see the 
actual JVM being used - but then I don't know how your package (or Debian?) 
abstracts the Java version mess.
| > | > | > 
| > | > | > This was vanilla 'openjdk-9-jdk' and nothing else.
| > | > | > 
| > | > | > Could I / should invoke 'R CMD javareconf' ?
| > | > | > 
| > | > | > Dirk
| > | > | > 
| > | > | > | Cheers,
| > | > | > | Simon
| > | > | > | 
| > | > | > | 
| > | > | > | > On Oct 17, 2017, at 10:24 PM, Dirk Eddelbuettel 
<e...@debian.org> wrote:
| > | > | > | > 
| > | > | > | > 
| > | > | > | > Simon,
| > | > | > | > 
| > | > | > | > With your new rJava_0.9-9 I tried this against Debian's 
openjdk-9-jdk -- but
| > | > | > | > with R 3.4.2 as built against openjdk-7 -- and it still fails:
| > | > | > | > 
| > | > | > | > checking whether setjmp.h is POSIX.1 compatible... yes
| > | > | > | > checking whether sigsetjmp is declared... yes
| > | > | > | > checking whether siglongjmp is declared... yes
| > | > | > | > checking Java support in R... present:
| > | > | > | > interpreter : '/usr/lib/jvm/default-java/jre/bin/java'
| > | > | > | > archiver    : '/usr/lib/jvm/default-java/bin/jar'
| > | > | > | > compiler    : '/usr/lib/jvm/default-java/bin/javac'
| > | > | > | > header prep.: '/usr/lib/jvm/default-java/bin/javah'
| > | > | > | > cpp flags   : '-I/usr/lib/jvm/default-java/include 
-I/usr/lib/jvm/default-java/include/linux'
| > | > | > | > java libs   : '-L/usr/lib/jvm/default-java/jre/lib/amd64/server 
-ljvm'
| > | > | > | > checking whether Java run-time works... ./configure: line 3736: 
/usr/lib/jvm/default-java/jre/bin/java: No such file or directory
| > | > | > | > no
| > | > | > | > configure: error: Java interpreter 
'/usr/lib/jvm/default-java/jre/bin/java' does not work
| > | > | > | > ERROR: configuration failed for package 'rJava'
| > | > | > | > * removing 
'/build/rjava-0.9-9/debian/r-cran-rjava/usr/lib/R/site-library/rJava'
| > | > | > | > /usr/share/R/debian/r-cran.mk:101: recipe for target 
'R_any_arch' failed
| > | > | > | > make: *** [R_any_arch] Error 1
| > | > | > | > dpkg-buildpackage: error: fakeroot debian/rules binary gave 
error exit status 2
| > | > | > | > 
| > | > | > | > 
| > | > | > | > Any ideas?  I saw that r-devel got some changes related to 
javareconf.  Do we
| > | > | > | > need to port that to r-patched?  Can you advise?
| > | > | > | > 
| > | > | > | > Many thanks,  Dirk
| > | > | > | > 
| > | > | > | > -- 
| > | > | > | > http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
| > | > | > | > 
| > | > | > | 
| > | > | > 
| > | > | > -- 
| > | > | > http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
| > | > | > 
| > | > | 
| > | > 
| > | > -- 
| > | > http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
| > | > 
| > | 
| > 
| > -- 
| > http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
| > 
| 

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org

Reply via email to