Make another attempt #1685
Make another attempt #1685
Conversation
|
There is one other issue I don't know how to solve. The command line parser stops on an unknown switch. So if the command line has anything that isn't -mca, -am, or -tune its possible not all the options will be parsed. I have a fix but it will potentially parse options to the app. I think that is probably fine. I will put together a patch and we can discuss it. |
|
@jjhursey How in the world do I find the error being reported by your tests? All I see is a mass of output, and I can't find the error. |
|
@hjelmn I took a couple of simple shots at it, but ran into infinite loops. So I'll have to defer until you have your patch available. Can you please set it up as a PR against this PR? |
|
Sure. I ran into the same issue. I will create a PR later today. |
|
@rhc54 Looking at the few tests that failed, the bottom of file looks like: #################
Run Examples
#################
+ cd /gpfs/gpfs_stage1/jhursey/jenkins/workspace/ompi_public_pr_master/ompi-src/examples
+ timeout --preserve-status -k 22s 20s mpirun -np 2 -mca btl tcp,vader,sm,self hello_c
#################
# Summary
#################
# Tests Passed: 5 of 7
#---------------------
IBM_CI_SUCCESS : Autogen
IBM_CI_SUCCESS : Configure
IBM_CI_SUCCESS : Make
IBM_CI_SUCCESS : Make Install
IBM_CI_SUCCESS : Make examples
#################When a CI test fails I give you all of the output generated from the whole build process, which is a lot. I put a summary at the bottom of what passed to help a bit with that. These seem to have failed in the test run - see the line below (the + timeout --preserve-status -k 22s 20s mpirun -np 2 -mca btl tcp,vader,sm,self hello_cThat must have timedout. I'll take a note to see if I can get it to print an error message when that happens before it fails the build. |
Maintain the python bindings Only call cmd_line_setup once as those params will persist Temporarily disable the AMCA option
|
@hjelmn I think we've made a good-faith attempt, but it's time to surrender pending identification of a maintainer for this code. I've added a check-and-abort so we don't silently ignore the AMCA option. Meantime, I'd suggest we consider this code "non-operational pending identification of a maintainer" ala what we did for checkpoint-restart. |
|
@hjelmn Github problems are preventing the code on the web page from updating - when it does, you'll see that I squashed it down to a single commit and added the check-and-abort logic to mca_base_var.c. |
This commit changes the command line parsing behavior when ignoring both unknown switches and tokens. In this case we are probably trying to parse part of the command line (-mca, -am, -tune) before the rest. Before this change the command line parser stopped on the first unknown switch or token. Now it will continue to parse out options. This includes options to the app being run by mpirun (unless -- is specified) but since only -mca, -tune, and -am are affected this probably is not a big deal. This change only affects the parsing of command-line MCA parameters in orte-submit, all other parsers are unchanged. Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
opal/cmd_line: enable complete command line parsing
… - put them in with the rest of the OPAL MCA param registrations Take another shot at untangling the spaghetti orterun: fix for command line parsing orte-submit calls opal_init_util () before parsing out MCA command line options (-mca, -am, etc). This prevents mpirun from setting opal MCA variables for some frameworks as well as the MCA base. This is because when a framework is opened all of its variables are set to read-only. Eventually we want to lift this restriction on some MCA variables but since -mca is affected we must parse out the MCA command line options before opal_init_util(). This commit fixes the bug by adding a new option to opal_cmd_line_parse (ignore unknown option) so orte-submit can pre-parse the command line for MCA options. Signed-off-by: Nathan Hjelm <hjelmn@me.com> Minor cleanups to avoid releasing/recreating the cmd line
|
See #1692 |
|
Closed in favor of other fix |
3 participants
No description provided.