Re: amber-developers: protocols for checking things in to the Amber 9 CVS tree

From: Scott Brozell <sbrozell.scripps.edu>
Date: Wed, 20 Apr 2005 16:55:38 -0700

Hi,

On Wed, 20 Apr 2005, David A. Case wrote:

> Hi everyone:
>
> We've had a flurry of development activity recently, which is good. But
also
> a number of (avoidable) glitches, which is bad.
>
> Everyone understands that code in the Amber9 tree is not all that
stable, and
> might have bugs in it. So I don't want people to be afraid to try out
new
> ideas.
>
> However, you should be sure that your revisions compile cleanly, and
pass the
> tests, or do not introduce new failures beyond those listed here:
>
> http://amber.scripps.edu/cgi-bin/wiki.pl/Amber_9_test_failures
>
> If your code changes the expected output of the tests, then you should
> generally update the test cases when you update the code. If you must

This has been advertised before, but as a refresher, there is a script
that takes the tedium out of updating the test save files:
test/update
Usage: ./update [-v] regular_expression

Description
       Search *.dif files for regular_expression and make the
       corresponding output files the .save files.

Options
       -v
              verbose.


Here is how to use it:
make test
make check
# Carefully examine the dif files to determine that the diffs are all
# expected and correct; then pick a suitably large regular expression
# that will match your changes.
./update -v the_regular_expression


I modified a couple .save files to create this example:

long ago bohr 007> /usr/local/srb/9/test make test
Apr 20 4:34:48pm bohr 557> /usr/local/srb/9/test make check
(find . -name "*.dif" ! -size 0c -print | \
while read dif ;\
do \
    echo $dif ;\
    cat $dif ;\
    echo ;\
done ;\
)
./dmp/mdout.dmp.dif
2c2
< Amber 8 SANDER Scripps/UCSF 2004

---
>           Amber 9  SANDER                 Scripps/UCSF 2004
./nonper/mdout.belly.dif
2c2
<           Amber 8  SANDER                 Scripps/UCSF 2004
---
>           Amber 9  SANDER                 Scripps/UCSF 2004
Apr 20  4:44:22pm bohr 572> /usr/local/srb/9/test  ./update -v 'Amber 8
SANDER'
Searching ./dmp/mdout.dmp.dif
<           Amber 8  SANDER                 Scripps/UCSF 2004
    match found; printing ./dmp/mdout.dmp.dif:
2c2
<           Amber 8  SANDER                 Scripps/UCSF 2004
---
>           Amber 9  SANDER                 Scripps/UCSF 2004
move mdout.dmp to mdout.dmp.save ?
y
Move successful
Searching ./nonper/mdout.belly.dif
<           Amber 8  SANDER                 Scripps/UCSF 2004
    match found; printing ./nonper/mdout.belly.dif:
2c2
<           Amber 8  SANDER                 Scripps/UCSF 2004
---
>           Amber 9  SANDER                 Scripps/UCSF 2004
move mdout.belly to mdout.belly.save ?
y
Move successful
Scott
> introduce new test failures, then please document that at the above
page,
> so that people will know to ignore them.
>
> If possible, please confirm that your code works with g95 as well as
ifort
> (this is for Fortran updates, obviously).  Since g95 is a stricter
compiler,
> that can help reduce problems down the line.
>
> When you introduce new features, introduce new test cases as well that
> exercise those features.  This will also help reduce future problems.
>
> If we all are a little more careful, I think we can save time for
everyone.
>
> ...thanks...dac
>

Received on Wed Apr 05 2006 - 23:49:57 PDT