mpich FTBFS on sh4: MPIUI_Thread TLS definition mismatches tbss

2017-06-06 Thread Drew Parsons
mpich has been consistently failing to build on sh4.
The error occurs when linking the static library with the message:

/usr/bin/ld: MPIUI_Thread: TLS definition in 
lib/.libs/libmpich.a(lib_libmpich_la-mpiu_thread.o) section .tbss mismatches 
non-TLS reference in lib/.libs/libmpich.a(lib_libmpich_la-comm_rank.o)
lib/.libs/libmpich.a: error adding symbols: Bad value

I'm not familiar with TLS or tbss.  Is there a straightforward way to
fix it?

The compilation of src/util/thread/mpiu_thread.c and
src/mpi/comm/comm_rank.c doesn't report anything odd.

The link error suggests 2 different versions of MPIUI_Thread are used. 
But I can only see a definition in src/util/thread/mpiu_thread.c.
src/mpi/comm does not refer to it, and comm_rank.c does not use any
MPIU object. So the error message doesn't make sense to me.

Can anyone make sense of it?


Drew



Bug#864288: openjdk-9: Please obsolete patch 8073754-stack-overflow-9-build.diff

2017-06-06 Thread John Paul Adrian Glaubitz
Source: openjdk-9
Version: 9~b170-2
Severity: important
User: debian-powerpc@lists.debian.org
Usertags: powerpc,powerpcspe,ppc64

Hi!

openjdk-9 currently fails to build from source on the powerpc* and ppc64
targets in Debian Ports because the minimum VM stack size is too small:

/<>/build-zero/jdk/bin/java -Xms64M -Xmx1600M 
-XX:ThreadStackSize=2240 -XX:+UseSerialGC -Xms32M -Xmx512M 
-XX:TieredStopAtLevel=1 -cp 
/<>/build-zero/buildtools/tools_jigsaw_classes --add-exports 
java.base/jdk.internal.module=ALL-UNNAMED 
build.tools.jigsaw.AddPackagesAttribute /<>/build-zero/jdk
Error: Could not create the Java Virtual Machine.
Error: A fatal exception has occurred. Program will exit.

The Java thread stack size specified is too small. Specify at least 2432k
ExplodedImageOptimize.gmk:40: recipe for target 
'/<>/build-zero/jdk/_packages_attribute.done' failed
make[4]: *** [/<>/build-zero/jdk/_packages_attribute.done] Error 1

>From the internal builds at SUSE I know that openjdk-9 builds just fine on 
>ppc64
without any modifications to the stack size and increasing the stack size in
the patch 8073754-stack-overflow-9-build.diff didn't help either.

So I just dropped the patch debian/patches/8073754-stack-overflow-9-build.diff
in the Debian package which made the build succeed on ppc64, it still fails
on powerpc for other reasons though. I didn't test powerpcspe but I assume it
should be the same, either way. After removing this patch, the patch
s390x-thread-stack-size.diff didn't apply anymore, so this patch needs to be
rebased.

Thanks for consideration!

Adrian

 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaub...@debian.org
`. `'   Freie Universitaet Berlin - glaub...@physik.fu-berlin.de
  `-GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913



Re: mpich FTBFS on sh4: MPIUI_Thread TLS definition mismatches tbss

2017-06-06 Thread Aaron M. Ucko
Drew Parsons  writes:

> The link error suggests 2 different versions of MPIUI_Thread are used. 
> But I can only see a definition in src/util/thread/mpiu_thread.c.
> src/mpi/comm does not refer to it, and comm_rank.c does not use any
> MPIU object. So the error message doesn't make sense to me.

comm_rank's reference to MPIUI_Thread is also present on at least amd64,
and presumably due to some macro that (directly or indirectly) calls
MPIU_THREADPRIV_FIELD.  However, it's not clear why multithreaded build
settings would be in effect for mpiu_thread but not also comm_rank.  (It
looks like MPIUI_Thread's thread-locality is conditional on
MPICH_IS_THREADED, which mpichconf.h defines centrally.)

I don't have time to dig deeper, but hope that brief analysis helps.

-- 
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?a...@monk.mit.edu