<div dir="ltr">src/ksp/ksp/tutorials/output/ex71_bddc_elast_both_approx_<b>alt.</b>out uses ML and src/ksp/ksp/tutorials/output/ex71_bddc_elast_both_approx.out uses GAMG.<div><br></div><div>The test seems to look at the alt file and not the normal one. I don't understand. I do get an error message ...</div><div><br></div><div>11:42 adams/cheby-spd-cg= ~/Codes/petsc-master$ make cleantest<br>/usr/bin/make  --no-print-directory -f gmakefile.test PETSC_ARCH=arch-macosx-gnu-g PETSC_DIR=/Users/markadams/Codes/petsc cleantest<br>/bin/rm -f -r ./arch-macosx-gnu-g/tests ./arch-macosx-gnu-g/tests/testfiles<br>11:42 adams/cheby-spd-cg= ~/Codes/petsc-master$ make -f gmakefile test search='ksp_ksp_tutorials-ex71_bddc_elast_both%' PETSC_DIR=$PWD<br><b>gmakefile.test:92: arch-macosx-gnu-g/tests/testfiles: No such file or directory<br></b>/System/Library/Frameworks/Python.framework/Versions/2.7/Resources/Python.app/Contents/MacOS/Python /Users/markadams/Codes/petsc-master/config/gmakegentest.py --petsc-dir=/Users/markadams/Codes/petsc-master --petsc-arch=arch-macosx-gnu-g --testdir=./arch-macosx-gnu-g/tests<br>Using MAKEFLAGS: PETSC_DIR=/Users/markadams/Codes/petsc-master search=ksp_ksp_tutorials-ex71_bddc_elast_both%<br>          CC arch-macosx-gnu-g/tests/ksp/ksp/tutorials/ex71.o<br>     CLINKER arch-macosx-gnu-g/tests/ksp/ksp/tutorials/ex71<br>        TEST arch-macosx-gnu-g/tests/counts/ksp_ksp_tutorials-ex71_bddc_elast_both_approx.counts<br> ok ksp_ksp_tutorials-ex71_bddc_elast_both_approx<br>not ok diff-ksp_ksp_tutorials-ex71_bddc_elast_both_approx # Error code: 1<br>#       1,13c1<br>#       <   0 KSP Residual norm 1615.07 <br># <   1 KSP Residual norm 420.868 <br># <   2 KSP Residual norm 187.45 <br>#  <   3 KSP Residual norm 67.3919 <br># <   4 KSP Residual norm 21.3237 <br># <   5 KSP Residual norm 5.8091 <br>#  <   6 KSP Residual norm 1.0923 <br>#  <   7 KSP Residual norm 0.527464 <br>#        <   8 KSP Residual norm 0.380684 <br>#        <   9 KSP Residual norm 0.0354163 <br>#       <  10 KSP Residual norm 0.0237308 <br>#       <  11 KSP Residual norm 0.0121289 <br>#       < Linear solve converged due to CONVERGED_RTOL iterations 11<br>#      ---<br>#  > Linear solve converged due to CONVERGED_RTOL iterations 10<br>#      77,92d64<br>#     <     PC Object: 1 MPI processes<br>#        <       type: shell<br>#    <         Nullspace corrected interior solve<br>#  <         L:<br>#  <           Mat Object: 1 MPI processes<br>#      <             type: seqdense<br>#        <             rows=144, cols=6<br>#      <             total: nonzeros=864, allocated nonzeros=864<br>#   <             total number of mallocs used during MatSetValues calls=0<br>#      <         K:<br>#  <           Mat Object: 1 MPI processes<br>#      <             type: seqdense<br>#        <             rows=144, cols=6<br>#      <             total: nonzeros=864, allocated nonzeros=864<br>#   <             total number of mallocs used during MatSetValues calls=0<br>#      <         inner preconditioner:<br>#       94,95c66,67<br>#  <             type: ml<br>#      <               type is MULTIPLICATIVE, levels=3 cycles=v<br>#  ---<br>#  >       type: gamg<br>#     >         type is MULTIPLICATIVE, levels=2 cycles=v<br>#   97a70,77<br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jun 15, 2020 at 10:15 AM Satish Balay <<a href="mailto:balay@mcs.anl.gov" target="_blank">balay@mcs.anl.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On Mon, 15 Jun 2020, Mark Adams wrote:<br>
<br>
> My pipeline is failing on ksp/ex71.c and it seems to be picking up an "alt"<br>
> version of the output.<br>
<br>
Hm - it does a diff with (basic, alt) files. If all diffs fail - then it prints a diff from one of them.<br>
<br>
<br>
> I tried REPLACE=1 and both output files seemed to<br>
> change. What is going on with these "alt" output files?<br>
<br>
I'm not sure how this works with alt files. I assumed it ignores alt<br>
files - and updates the primary file.  Usually I would need a new alt<br>
file - so I just move this over manually to a new one [and keep the<br>
current files unchanged]<br>
<br>
Note: To test the new alt file - one need to do 'make cleantest' and<br>
rerun the test - otherwise the test harness does not know that it<br>
should pick up the new alt file.<br>
<br>
Satish<br>
</blockquote></div>