Hi Wei-keng,<br><br>I've found the problem - the type of the 4th variable in my call to nfmpi_get_var_all was default integer and should have been MPI_OFFSET_KIND. An f90 interface would have caught this...<br><br>Thanks for your help,<br>
<br>Jim<br><br><div class="gmail_quote">On Tue, Dec 29, 2009 at 11:20 AM, Wei-keng Liao <span dir="ltr"><<a href="mailto:wkliao@ece.northwestern.edu">wkliao@ece.northwestern.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi, Jim,<br>
<br>
I would like to mention that ncmpi_get_var_all() is to read the<br>
entire variable. The checking at line 2009 is to see if the user<br>
buffer has the same number of array elements as the entire<br>
variable. However, this is a warning, not an error. Please let<br>
me know if you are also getting incorrect data from this call.<br>
<br>
What puzzles me is the 2nd printf statement that the same<br>
nelems and cnelems bypass the if statement for checking<br>
(nelems != cnelems). Such problem may be because of mismatched<br>
pnetcdf object files in the library). Since you had a clean<br>
rebuild, this should not be the cause. Something else is fishy.<br>
<br>
If you are using the SVN most recent revision 753, the two<br>
printf statements you added should be in lines 2009 and 2013.<br>
Did you also add other codes above the line 2009?<br><font color="#888888">
<br>
<br>
Wei-keng</font><div><div></div><div class="h5"><br>
<br>
On Dec 29, 2009, at 9:58 AM, Jim Edwards wrote:<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
I rebuilt pnetcdf with all optimization off and still get the error.<br>
I tried to build a short testcase and it works fine, (no error - line 2016 is not printed).<br>
<br>
<br>
<br>
<br>
<br>
On Mon, Dec 28, 2009 at 6:45 PM, Wei-keng Liao <<a href="mailto:wkliao@ece.northwestern.edu" target="_blank">wkliao@ece.northwestern.edu</a>> wrote:<br>
<br>
Your outputs said both nelems and cnelems are 27.<br>
I don't know why the printf at line 2016 even got printed.<br>
Can you try a clean build of pnetcdf?<br>
<br>
Or if you can provide a test code, I can give it a try.<br>
<br>
<br>
Wei-keng<br>
<br>
<br>
On Dec 28, 2009, at 6:59 PM, Jim Edwards wrote:<br>
<br>
That works - thanks. Here is another one. I'm gettiing an error<br>
Input/Output data amount mismatch<br>
<br>
that I've traced to the mpinetcdf.c file at around line 2000 where I added a couple of print statements<br>
<br>
<br>
printf("line %d nelems %d cnelems %d\n",__LINE__,nelems,cnelems);<br>
if (nelems != cnelems) {<br>
if (warning == NC_NOERR)<br>
warning = NC_EIOMISMATCH;<br>
printf("line %d nelems %d cnelems %d\n",__LINE__,nelems,cnelems);<br>
<br>
(nelems>cnelems) ? (nelems=cnelems) : (cnelems=nelems);<br>
}<br>
<br>
<br>
This results in the output:<br>
<br>
1:line 2012 nelems 27 cnelems 27<br>
1:line 2016 nelems 27 cnelems 27<br>
<br>
<br>
<br>
<br>
<br>
<br>
On Mon, Dec 28, 2009 at 5:38 PM, Wei-keng Liao <<a href="mailto:wkliao@ece.northwestern.edu" target="_blank">wkliao@ece.northwestern.edu</a>> wrote:<br>
Hi,<br>
<br>
If you run autoreconf first, it will generate that file for you.<br>
<br>
Wei-keng<br>
<br>
<br>
<br>
On Dec 28, 2009, at 2:47 PM, Jim Edwards wrote:<br>
<br>
Hi,<br>
<br>
Tried checking out the svn trunk and get the following configure error...<br>
<br>
configure: error: cannot find sources (src/lib/<a href="http://ncconfig.in" target="_blank">ncconfig.in</a>) in . or ..<br>
<br>
what happened to this file?<br>
<br>
<br>
<br>
<br>
</blockquote>
<br>
</div></div></blockquote></div><br>