[MOAB-dev] ParallelComm.cpp - potential bug
Vijay S. Mahadevan
vijay.m at gmail.com
Tue Dec 4 18:42:40 CST 2012
I pulled the latest moab source and compiled with Intel on a linux machine.
I get the following warning which could be serious:
ParallelComm.cpp(4091): remark #1599: declaration hides variable "k"
(declared at line 4068)
variable k is defined at both 4068 and 4091. I would've modified the
inner loop to use a different variable but am not sure how to modify
the source inside correspondingly. Can someone please fix this since
it could cause some serious bugs because the variable 'k' is ambiguous
in lines 4094:4098. Using some descriptive variable names for indexes
would also be helpful rather than i, j, k etc..
4068 for (int k=0; k<nbPairsMultiShared; k++)
4069 {
4070 int startMulti=multiSharedStartEnd[2*k];
4071 int endMulti =multiSharedStartEnd[2*k+1];
4072 // so tuples indices start, end, will be remote sent for multi sharing
4073 // first tuple sent will be with the owner infos, the other
4074 for (int j=startMulti; j<=endMulti; j++)
4075 {
4076 // first send to processor the info about the local handle
(which will be the owner)
4077 int proc_to_send_to = remoting_entities.vi_wr[j];
4078 EntityHandle handle_on_owning_proc =
remoting_entities.vul_wr[2*j+1]; // it was remote before
...................
...................
4089 multishared_verts.inc_n();
4090 // create more tuples, for each instance
4091 for (int k=startMulti; k<=endMulti; k++)
4092 {
4093 multishared_verts.vi_wr[2*indexInTuple] =
proc_to_send_to; // send to
4094 multishared_verts.vi_wr[2*indexInTuple+1] =
remoting_entities.vi_wr[k]; // multi proc
4095 multishared_verts.vl_wr[indexInTuple] =
remoting_entities.vl_wr[k]; // gid, for debug
4096 multishared_verts.vul_wr[2*indexInTuple] =
handle_with_multi; // remote handle on proc_to_send_to
4097 multishared_verts.vul_wr[2*indexInTuple+1] =
remoting_entities.vul_wr[2*k]; // handle on multi proc
4098 indexInTuple++;
4099 multishared_verts.inc_n();
4100 }
4101 }
4102 }
Vijay
More information about the moab-dev
mailing list