As always, Paul, we deeply appreciate your "torture" sessions. They are invaluable for turning up problems (e.g., the entire Fortran rathole of the last few days).

Your methods are just fine - the problem with issues being fixed in the branch and not the trunk is entirely on our side. People are just not being careful about following the procedures, and that's something we need to straighten out to avoid these problems in the future.

We'll take you up on your offer to add you to the cc list where appropriate - will try not to overload you too much!

Thanks again

On Jan 24, 2014, at 2:04 PM, Paul Hargrove <> wrote:

It looks like, with the possible exception of iterating with Jeff on Fortran configure tests, my (annual?) round of torture-the-open-mpi-dev^H^H^Hsources is nearing its end.

Some of have probably noticed that I started mining the ticket system for things to test.  I've searched for my name, for certain platforms I have access to, and have skimmed all the open CMRs and many defects for v1.7.  I've added myself to the CC field on some where I may have the (unique?) ability to verify that the fix was effective.

I did not search closed tickets.  If there are any (non-documentation) issues for 1.7.4 that have been closed, then my testing last night PROBABLY covered them.  However, if there are specific configure options (like --enable-builtin-atomics for #3081) then I probably did NOT test it last night.  So, if there are any items in v1.7 that you would specifically like me to (re)test, please let me know (CCing user "phargrov" in trac is preferred if a ticket already exists).

For the future, when there are tickets for defects where I was the original (devel list?) reporter OR you think I may be the best/only candidate to verify a fix, please feel free to add me to the CC field.  At worse I'll ignore the emails, but at best I can (re)test to confirm that a fix (either proposed or committed) actually works.  By adding me to the CC field, a portion of this process can continue year-round instead of being all crammed into the RC period.  This applies ESPECIALLY to CMRs to 1.7.5, to avoid unnecessary delays of v1.8.

An aside:
At least twice in this round of testing it was discovered that issues I had reported in 1.5/1.6 got fixed in the branch WITHOUT being fixed in the trunk.  If there is some flaw in the process that could be addressed by changing how *I* operate, please let me know.  I am hesitant to switch from mailing-list reports to entering my own tickets, but could if that it preferred.


Paul H. Hargrove                
Future Technologies Group
Computer and Data Sciences Department     Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900
devel mailing list