[mvapich-discuss] Patches to fix v1.9rc1 compiler warnings

[Jonathan Perkins]

Thanks for the patches Adam.  We'll take a look at these before our
final release.

ch3:mrail is still the preferred interface.

On Fri, Apr 26, 2013 at 05:52:47PM -0700, Adam T. Moody wrote:
> Hi Jonathan,
> Here's another under the ch3:mrail/with-rdma=gen2 build:
> 
> src/mpid/ch3/src/mpid_abort.c: In function 'MPID_Abort':
> src/mpid/ch3/src/mpid_abort.c:93: warning: passing argument 2 of
> 'MPIDI_CH3_Abort' discards qualifiers from pointer target type
> ./src/mpid/ch3/include/mpidimpl.h:1761: note: expected 'char *' but
> argument is of type 'const char *'
> 
> -Adam
> 
> 
> Adam T. Moody wrote:
> 
> >Hi Jonathan,
> >I get some warnings when building 1.9rc1 for PSM.
> >
> >The allocate_buffer.patch fixes the following:
> >
> >osu_gather.c: In function 'main':
> >osu_gather.c:75: warning: passing argument 1 of 'allocate_buffer'
> >from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >osu_gather.c:84: warning: passing argument 1 of 'allocate_buffer'
> >from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >osu_scatter.c: In function 'main':
> >osu_scatter.c:75: warning: passing argument 1 of 'allocate_buffer'
> >from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >osu_scatter.c:84: warning: passing argument 1 of 'allocate_buffer'
> >from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >osu_alltoall.c: In function 'main':
> >osu_alltoall.c:74: warning: passing argument 1 of
> >'allocate_buffer' from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >osu_alltoall.c:82: warning: passing argument 1 of
> >'allocate_buffer' from incompatible pointer type
> >osu_coll.h:273: note: expected 'void **' but argument is of type
> >'char **'
> >
> >
> >The lustre_warnings.patch fixes these:
> >
> >adio/ad_lustre/ad_lustre.c:17: warning: initialization from
> >incompatible pointer type
> >adio/ad_lustre/ad_lustre.c:19: warning: initialization from
> >incompatible pointer type
> >adio/ad_lustre/ad_lustre.c:24: warning: initialization from
> >incompatible pointer type
> >
> >
> >The psm_gnu_source.patch fixes:
> >
> >In file included from ./src/mpid/ch3/include/mpidimpl.h:1926,
> >                from ./src/include/coll_shmem.h:25,
> >                from src/mpi/coll/ch3_shmem_coll.c:44:
> >./src/mpid/ch3/channels/psm/include/psmpriv.h:16:1: warning:
> >"_GNU_SOURCE" redefined
> >src/mpi/coll/ch3_shmem_coll.c:21:1: warning: this is the location
> >of the previous definition
> >
> >
> >I still get the following, but there are two ways you might fix
> >this and I don't know which you'd prefer:
> >
> >src/mpid/ch3/channels/psm/src/psm_entry.c: In function 'psm_doinit':
> >src/mpid/ch3/channels/psm/src/psm_entry.c:113: warning: passing
> >argument 1 of 'MPIDI_CH3U_Comm_register_create_hook' from
> >incompatible pointer type
> >./src/mpid/ch3/include/mpidimpl.h:1706: note: expected 'int
> >(*)(struct MPID_Comm *, void *)' but argument is of type 'int
> >(*)(struct MPID_Comm *)'
> >src/mpid/ch3/channels/psm/src/psm_entry.c:202: warning: passing
> >argument 1 of 'MPIDI_CH3U_Comm_register_destroy_hook' from
> >incompatible pointer type
> >./src/mpid/ch3/include/mpidimpl.h:1707: note: expected 'int
> >(*)(struct MPID_Comm *, void *)' but argument is of type 'int
> >(*)(struct MPID_Comm *)'
> >
> >
> >Also, when building with the ch3:nemesis:ib channel, I get the following:
> >
> >src/mpid/ch3/channels/nemesis/netmod/ib/ib_vc.c: In function
> >'MPID_nem_ib_vc_init':
> >src/mpid/ch3/channels/nemesis/netmod/ib/ib_vc.c:182: warning:
> >assignment from incompatible pointer type
> >
> >
> >Is the ch3:nemesis:ib channel now recommended over the
> >ch3:mrail/with-rdma=gen2 channel for Mellanox systems?
> >-Adam
> >
> >
> >Jonathan Perkins wrote:
> >
> >>On Fri, Apr 26, 2013 at 11:42:21AM -0700, Adam T. Moody wrote:
> >>
> >>>Hello MVAPICH team,
> >>>Here is a patch to v1.7 that I'd like to add to v1.9.  It fixes a
> >>>bug on PSM systems in which an MPI job is started from another,
> >>>e.g., a python+MPI test framework that mpiruns lots of other MPI
> >>>jobs as test cases.
> >>>
> >>>https://github.com/hpc/mvapich2-cce/commit/44c904359fc2325efc7b282998264d1a73a7639a
> >>>
> >>
> >>
> >>Thanks for the patch Adam.  The patch is straightforward and will be
> >>present in the 1.9 release.
> >>
> >>
> >------------------------------------------------------------------------
> >
> >_______________________________________________
> >mvapich-discuss mailing list
> >mvapich-discuss at cse.ohio-state.edu
> >http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
> >
> 

-- 
Jonathan Perkins
http://www.cse.ohio-state.edu/~perkinjo