Skip to content

IBM tool/init_after_finalize failures #9013

Closed
@hppritcha

Description

@hppritcha

I am seeing asserts with the IBM tool/init_after_finalize test of the pattern

closing opal framework btl  <------------ my debug statement
init_after_finalize: mca_base_var_group.c:375: mca_base_var_group_deregister: Assertion `((0xdeafbeedULL << 32) + 0xdeafbeedULL) == ((opal_object_t *) (enums[i]))->obj_magic_id' failed.
[st-master:28398] *** Process received signal ***
[st-master:28398] Signal: Aborted (6)
[st-master:28398] Signal code:  (-6)
[st-master:28398] [ 0] [0xffffbd43066c]
[st-master:28398] [ 1] /lib64/libc.so.6(gsignal+0x4c)[0xffffbcd950e8]
[st-master:28398] [ 2] /lib64/libc.so.6(abort+0x11c)[0xffffbcd96760]
[st-master:28398] [ 3] /lib64/libc.so.6(+0x2e5dc)[0xffffbcd8e5dc]
[st-master:28398] [ 4] /lib64/libc.so.6(__assert_perror_fail+0x0)[0xffffbcd8e694]
[st-master:28398] [ 5] /users/hpritchard/ompi/install_master_ucx/lib/libopen-pal.so.0(mca_base_var_group_deregister+0x1f4)[0xffffbcc74378]
[st-master:28398] [ 6] /users/hpritchard/ompi/install_master_ucx/lib/libopen-pal.so.0(mca_base_var_group_deregister+0x348)[0xffffbcc744cc]
[st-master:28398] [ 7] /users/hpritchard/ompi/install_master_ucx/lib/libopen-pal.so.0(mca_base_framework_close+0x114)[0xffffbcc76858]
[st-master:28398] [ 8] /users/hpritchard/ompi/install_master_ucx/lib/libopen-pal.so.0(opal_info_close_components+0xbc)[0xffffbcc27448]
[st-master:28398] [ 9] /users/hpritchard/ompi/install_master_ucx/lib/libmpi.so.0(ompi_info_close_components+0xb0)[0xffffbcff3b4c]
[st-master:28398] [10] /users/hpritchard/ompi/install_master_ucx/lib/libmpi.so.0(PMPI_T_finalize+0x4c)[0xffffbd0a0c08]
[st-master:28398] [11] ./init_after_finalize[0x401174]

when compiling master with the following config options:

./configure --with-ucx=path_to_ucx_install_v1.10.0  --with-ofi=no --enable-debug
./configure --with-ucx=path_to_ucx_install_v1.10.0  --with-ofi=no --enable-debug --enable-mca-dso
./configure --with-ofi=path_to_ofi_install_1.12.1 --with-ucx=no --enable-debug
./configure --with-ofi=path_to_ofi_install_1.12.1 --with-ucx=no --enable-debug --enable-mca-dso

test works with the v4.1.x branch. presumably works on v4.0.x too.

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions