I just implemented a 'fix' for this in which if the client submits a
test_build or test_run without proper lineage instead of guessing
(any mostly getting it wrong) we just point the result to an 'undef'
row. This will allow us to more effectively track these going
forward, and maybe even remove them if needed.
So I guess I wouldn't worry about it past this morning, but it is
something we might want to ponder about a bit more by looking at how
the reporter/submit scripts should deal with Already Installed
scenarios.
-- Josh
On Sep 19, 2007, at 10:08 AM, Ethan Mallove wrote:
On Mon, Sep/17/2007 10:46:45AM, Josh Hursey wrote:
I've been getting quite a few errors from submit.php of
the below form. It seems that the MPI Install that is
being referenced is not valid. Could this is an 'already
installed' issue where the user is trying to submit
results to the database without having submitted an
mpi_install?
You are correct. Some test reporter-less runs were done
before some runs that used the MTTDatabase reporter were
done. I guess the rule of thumb is to, from the outset, run
using --trial if you plan to eventually send to the
database.
-Ethan
-- Josh
On Sep 17, 2007, at 10:37 AM, jjhur...@open-mpi.org wrote:
find_mpi_install_id():
The following SELECT returned -1:
-------------------------------
SELECT mpi_install_id
FROM mpi_install NATURAL JOIN
mpi_get NATURAL JOIN
compiler NATURAL JOIN
compute_cluster NATURAL JOIN
submit
WHERE
mpi_version = '1.2.4b1r16122M' AND
mpi_name = 'developer' AND
compiler_version = '4.1.2' AND
hostname = 'fog01' AND
mtt_client_version = '2.1devel' AND
local_username = 'paklui' AND
platform_name = 'fog01'
ORDER BY mpi_install_id DESC limit 1
-------------------------------
_______________________________________________
mtt-devel mailing list
mtt-de...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
_______________________________________________
mtt-devel mailing list
mtt-de...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/mtt-devel
