[cig-commits] commit 2331 by bangerth to /var/svn/dealii/aspect
dealii.demon at gmail.com
dealii.demon at gmail.com
Wed Mar 5 09:03:31 PST 2014
Revision 2331
Improve the way we print a warning message.
U trunk/aspect/source/postprocess/visualization.cc
http://www.dealii.org/websvn/revision.php?repname=Aspect+Repository&path=%2F&rev=2331&peg=2331
Diff:
Modified: trunk/aspect/source/postprocess/visualization.cc
===================================================================
--- trunk/aspect/source/postprocess/visualization.cc 2014-03-05 15:57:33 UTC (rev 2330)
+++ trunk/aspect/source/postprocess/visualization.cc 2014-03-05 17:03:28 UTC (rev 2331)
@@ -1,5 +1,5 @@
/*
- Copyright (C) 2011, 2012, 2013 by the authors of the ASPECT code.
+ Copyright (C) 2011, 2012, 2013, 2014 by the authors of the ASPECT code.
This file is part of ASPECT.
@@ -447,16 +447,22 @@
tmp_filename = tmp_filename_x;
delete []tmp_filename_x;
- // If we failed to create the temp file, just write directly to the target file
+ // If we failed to create the temp file, just write directly to the target file.
+ // We also provide a warning about this fact. There are places where
+ // this fails *on every node*, so we will get a lot of warning messages
+ // into the output; in these cases, just writing multiple pieces to
+ // std::cerr will produce an unreadable mass of text; rather, first
+ // assemble the error message completely, and then output it atomically
if (tmp_file_desc == -1)
{
- std::cerr << "***** WARNING: could not create temporary file, will "
- "output directly to final location. This may negatively "
- "affect performance."
- " (On processor "
- << Utilities::MPI::this_mpi_process (MPI_COMM_WORLD) << ".)"
- << std::endl;
+ std::string x ("***** WARNING: could not create temporary file, will "
+ "output directly to final location. This may negatively "
+ "affect performance."
+ " (On processor ");
+ x += Utilities::MPI::this_mpi_process (MPI_COMM_WORLD) + ".)
";
+ std::cerr << x << std::flush;
+
tmp_filename = *filename;
}
}
More information about the CIG-COMMITS
mailing list