<div>Many thanks for your precious feedback.</div><div>My MPICH install is 64 bit but it seems that for some reason 32 bit VC compiler is invoked (I have both 32 and 64 installed).</div><div>I am clarifying with a local MSVC expert how to control this, will update you later today.</div>
<div>Dominik</div><div><div><br><div class="gmail_quote">On Thu, Jul 22, 2010 at 3:45 PM, Satish Balay <span dir="ltr">&lt;<a href="mailto:balay@mcs.anl.gov">balay@mcs.anl.gov</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">&gt; sh: Warning: win32fe: File Not Found: /cygdrive/c/Program<br>
&gt; Files/MPICH2/lib/amd64/msmpi.lib<br>
<br>
</div>Looks like - if win32fe doesn&#39;t find the file - it doesn&#39;t try to<br>
convert the path. This is not the cause of his configure failure.<br>
<br>
The primary issue is: [which is much ahead in configure.log]<br>
<br>
&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;<br>
sh: /cygdrive/c/Users/Dominik/Programs/petsc-3.1-p3/bin/win32fe/win32fe cl  -o conftest.exe    -MT -wd4996  conftest.o  /cygdrive/c/Program\ Files/MPICH2/lib/mpi.lib Ws2_32.lib<br>
Executing: /cygdrive/c/Users/Dominik/Programs/petsc-3.1-p3/bin/win32fe/win32fe cl  -o conftest.exe    -MT -wd4996  conftest.o  /cygdrive/c/Program\ Files/MPICH2/lib/mpi.lib Ws2_32.lib<br>
sh: conftest.obj : error LNK2019: unresolved external symbol _MPI_Init referenced in function _main^M<br>
C:\Users\Dominik\Programs\PETSC-~1.1-P\conftest.exe : fatal error LNK1120: 1 unresolved externals^M<br>
&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;<br>
<br>
Here cl is finding the correct library - but not MPI_Init symbol in it.<br>
For whatever reason - libmpi.lib is not compatible.<br>
<font color="#888888"><br>
Satish<br>
</font><div><div></div><div class="h5"><br>
On Thu, 22 Jul 2010, Farshid Mossaiby wrote:<br>
<br>
&gt; I doubt &#39;cl&#39; can interpret Cygwin paths:<br>
&gt;<br>
&gt; cl : Command line warning D9002 : ignoring unknown option<br>
&gt; &#39;/cygdrive/c/Program Files/MPICH2/lib/amd64/msmpi.lib&#39;<br>
&gt;<br>
&gt; Shouldn&#39;t win32fe convert this path first?<br>
&gt;<br>
&gt; Hope this helps.<br>
&gt;<br>
&gt; --- On Wed, 7/21/10, Dominik Szczerba &lt;<a href="mailto:dominik@itis.ethz.ch">dominik@itis.ethz.ch</a>&gt; wrote:<br>
&gt;<br>
&gt; &gt; From: Dominik Szczerba &lt;<a href="mailto:dominik@itis.ethz.ch">dominik@itis.ethz.ch</a>&gt;<br>
&gt; &gt; Subject: Re: [petsc-users] configure problems<br>
&gt; &gt; To: &quot;PETSc users list&quot; &lt;<a href="mailto:petsc-users@mcs.anl.gov">petsc-users@mcs.anl.gov</a>&gt;<br>
&gt; &gt; Date: Wednesday, July 21, 2010, 7:01 PM<br>
&gt; &gt; Coming back to the original problem:<br>
&gt; &gt; I tried to compile MPICH2 myself<br>
&gt; &gt; and failed because of wrong version of automake in Cygwin.<br>
&gt; &gt; I then<br>
&gt; &gt; installed the Windows binary available on the mpich2<br>
&gt; &gt; webpage. I<br>
&gt; &gt; configure petsc with:<br>
&gt; &gt;<br>
&gt; &gt; $ ./config/configure.py PETSC_DIR=$PWD<br>
&gt; &gt; PETSC_ARCH=win64-msvc-release<br>
&gt; &gt; --download-c-blas-lapack=1<br>
&gt; &gt; --with-mpi-dir=&#39;/cygdrive/c/Program<br>
&gt; &gt; Files/MPICH2&#39; --with-x=0 --with-debugging=0<br>
&gt; &gt; --with-cc=&#39;win32fe cl&#39;<br>
&gt; &gt; --with-fc=0<br>
&gt; &gt;<br>
&gt; &gt; to hear at the end (of a very long configuration<br>
&gt; &gt; process...):<br>
&gt; &gt;<br>
&gt; &gt; sh:<br>
&gt; &gt; /cygdrive/c/Users/Dominik/Programs/petsc-3.1-p3/bin/win32fe/win32fe<br>
&gt; &gt; cl  -o conftest.exe    -MT -wd4996 <br>
&gt; &gt; conftest.o  /cygdrive/c/Program\<br>
&gt; &gt; Files/MPICH2/lib/amd64/msmpi.lib Ws2_32.lib<br>
&gt; &gt; Executing:<br>
&gt; &gt; /cygdrive/c/Users/Dominik/Programs/petsc-3.1-p3/bin/win32fe/win32fe<br>
&gt; &gt; cl  -o conftest.exe    -MT -wd4996 <br>
&gt; &gt; conftest.o  /cygdrive/c/Program\<br>
&gt; &gt; Files/MPICH2/lib/amd64/msmpi.lib Ws2_32.lib<br>
&gt; &gt; sh: Warning: win32fe: File Not Found: /cygdrive/c/Program<br>
&gt; &gt; Files/MPICH2/lib/amd64/msmpi.lib<br>
&gt; &gt; cl : Command line warning D9002 : ignoring unknown option<br>
&gt; &gt; &#39;/cygdrive/c/Program Files/MPICH2/lib/amd64/msmpi.lib&#39;<br>
&gt; &gt; conftest.obj : error LNK2019: unresolved external symbol<br>
&gt; &gt; _MPI_Init<br>
&gt; &gt; referenced in function _main<br>
&gt; &gt; C:\Users\Dominik\Programs\PETSC-~1.1-P\conftest.exe : fatal<br>
&gt; &gt; error<br>
&gt; &gt; LNK1120: 1 unresolved externals<br>
&gt; &gt;<br>
&gt; &gt; Any directions why would the configuration look for a<br>
&gt; &gt; nonexistent<br>
&gt; &gt; library? The contents of the installation folder are:<br>
&gt; &gt;<br>
&gt; &gt; $ ls -la /cygdrive/c/Program\ Files/MPICH2/lib/<br>
&gt; &gt; total 808<br>
&gt; &gt; drwx------+ 1 SYSTEM SYSTEM   4096<br>
&gt; &gt; 2010-07-21 15:54 .<br>
&gt; &gt; drwx------+ 1 SYSTEM SYSTEM   4096<br>
&gt; &gt; 2010-07-21 15:54 ..<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM   4644<br>
&gt; &gt; 2010-02-22 17:13 TraceInput.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM 322820 2010-02-22 17:09<br>
&gt; &gt; cxx.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM 131316 2010-02-22 17:11<br>
&gt; &gt; fmpich2.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM 136382 2010-02-22 17:13<br>
&gt; &gt; fmpich2g.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM   1936<br>
&gt; &gt; 2010-02-22 17:13 irlog2rlog.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM  10430 2010-02-22 16:30<br>
&gt; &gt; mpe.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM 132128 2010-02-22 16:29<br>
&gt; &gt; mpi.lib<br>
&gt; &gt; -rwx------+ 1 SYSTEM SYSTEM  61286 2010-02-22 16:30<br>
&gt; &gt; rlog.lib<br>
&gt; &gt;<br>
&gt; &gt; Best regards,<br>
&gt; &gt; Dominik<br>
&gt; &gt;<br>
&gt; &gt; On Tue, Jul 20, 2010 at 12:44 AM, Jed Brown &lt;<a href="mailto:jed@59a2.org">jed@59a2.org</a>&gt;<br>
&gt; &gt; wrote:<br>
&gt; &gt; &gt; On Tue, 20 Jul 2010 08:37:18 +1000, Matthew Knepley<br>
&gt; &gt; &lt;<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>&gt;<br>
&gt; &gt; wrote:<br>
&gt; &gt; &gt;&gt; This would do you absolutely no good. Even Jed&#39;s<br>
&gt; &gt; thing needs all the<br>
&gt; &gt; &gt;&gt; information that the configure provides. CMake is<br>
&gt; &gt; exactly what it<br>
&gt; &gt; &gt;&gt; says, make. It does no configuration<br>
&gt; &gt; &gt;<br>
&gt; &gt; &gt; This isn&#39;t true, CMake isn&#39;t a build system at all, it<br>
&gt; &gt; does<br>
&gt; &gt; &gt; configuration and produces build files for another<br>
&gt; &gt; tool (like make or an<br>
&gt; &gt; &gt; IDE).  I see no point in maintaining duplicate<br>
&gt; &gt; configuration systems,<br>
&gt; &gt; &gt; and CMake&#39;s scripting language is so attrocious that<br>
&gt; &gt; &quot;replacing&quot;<br>
&gt; &gt; &gt; BuildSystem with CMake script would be a disaster.<br>
&gt; &gt;  It&#39;s<br>
&gt; &gt; &gt; cross-compilation support is also a bit weak, and<br>
&gt; &gt; there are a pile of<br>
&gt; &gt; &gt; other things that it does poorly (mostly related to<br>
&gt; &gt; pathologically<br>
&gt; &gt; &gt; dysfunctional &quot;Find*.cmake&quot; modules).  That doesn&#39;t<br>
&gt; &gt; mean it isn&#39;t useful<br>
&gt; &gt; &gt; for producing build files.<br>
&gt; &gt; &gt;<br>
&gt; &gt; &gt; Jed<br>
&gt; &gt; &gt;<br>
&gt; &gt; &gt;<br>
&gt; &gt;<br>
&gt;<br>
&gt;<br>
&gt;<br>
&gt;<br>
</div></div></blockquote></div><br></div></div>