mpirun (master branch) error message for unknown and deprecated argument is wrong #10097

wzamazon · 2022-03-08T23:51:46Z

Thank you for taking the time to submit an issue!

Background information

What version of Open MPI are you using? (e.g., v3.0.5, v4.0.2, git branch name and hash, etc.)

master

Describe how Open MPI was installed (e.g., from a source/distribution tarball, from a git clone, from an operating system distribution package, etc.)

git clone from source. then configured with ./configure --prefix=/fsx/ALinux2/PortaFiducia/libraries/openmpi/master/install --with-hwloc=external --disable-man-pages

If you are building/installing from a git clone, please copy-n-paste the output from `git submodule status`.

94169f1725a26a156fd73fca1b61df14ccd581ab 3rd-party/openpmix (v1.1.3-3429-g94169f17)
 6722bd767e0fccbe07193f59f4f13e5610c7dd6a 3rd-party/prrte (psrvr-v2.0.0rc1-4274-g6722bd767e)

Please describe the system on which you are running

Operating system/version: Amazon Linux 2
Computer hardware: Intel
Network type: EFA

Details of the problem

The error messages of mpirun from master branch for deprecated/unknown argument is wrong.

For example:

If I pass a deprecated argument like -machinefile

/fsx/ALinux2/PortaFiducia/libraries/openmpi/master/install/bin/mpirun --map-by ppr:4:node -machinefile hostfile ./t

I got

--------------------------------------------------------------------------
An unrecognized option was included on the (null) command line:

  Option: ppr:4:node

Please use the "(null) --help" command to obtain a list of all
supported options.
--------------------------------------------------------------------------
prterun: command line error (Out of resource)

This is super confusing because "ppr:4:node" is valid input for --map-by. It is -machinefile that is causing trouble. So there seems to be an off-by-1 error in printing the arugment.

On the other hand, If I pass an unknown argument like -i like:

/fsx/ALinux2/PortaFiducia/libraries/openmpi/master/install/bin/mpirun -i

I got

--------------------------------------------------------------------------
An unrecognized option was included on the (null) command line:

  Option: -i

Please use the "(null) --help" command to obtain a list of all
supported options.
--------------------------------------------------------------------------

The error message is more helpful, though the (null) should be fixed.

The text was updated successfully, but these errors were encountered:

awlauria · 2022-03-09T15:35:42Z

Working on a fix. Should post this morning

awlauria · 2022-03-09T17:36:53Z

Fix for unknown argument "(null)":

openpmix/prrte#1257
openpmix/openpmix#2495

awlauria · 2022-03-09T18:58:01Z

Regarding the deprecation warning, it appears you are missing a - on -machinefile, which is causing the parser to get a bit wonky.

it seems to work when adding a double-dash.

FWIW it doesn't seem to be in prrte's deprecated list. I don't think it is deprecated.

We could add single-dash machinefile to the list of those converted from single-dash to double dash (or convert all of them..). I feel like prrte used to do that, but it got dropped at some point?

wzamazon · 2022-03-11T14:30:08Z

it appears you are missing a - on -machinefile, which is causing the parser to get a bit wonky.

Ah, I see. Thank you!

However, single dash is accepted by open mpi 4, so I think it should be handled by v5 (or at least print a proper warning).

We could add single-dash machinefile to the list of those converted from single-dash to double dash (or convert all of them..). I feel like prrte used to do that, but it got dropped at some point?

Yes, I remember prrte used to do that.

awlauria · 2022-03-11T14:46:28Z

No worries, I can try posting that patch, but may not until next week.

The fixes for "(null)" are fixed in pmix/prte master, just need to update the submodule.

wzamazon · 2022-03-11T15:40:31Z

I can confirm the "(null)" has been fixed by latest pmix/prte. Thank you!

This will preserve backwards compatability with the ompi v4 series with v5 and current master. Refs open-mpi/ompi#10097 Signed-off-by: Austen Lauria <awlauria@us.ibm.com>

This will preserve backwards compatability with the ompi v4 series with v5 and current master. First, this restores the silent conversion of all single dash args to multi-dash arguments. Second, it will track and eventually print all offending ompi args in a show-help message. It does this by appending all single dash arguments to an array, along with their position found in the arg list. This is key, since we don't really knoew where the user executable is on the list, and it isn't easy to tell. Luckily, pmix_arg_parse() will store where the ompi args stops in results->tail. Therefore all we need to know is if each offending argument came before results->tail or not to decide if a warning should be printed for it. This will prevent prte from erronously warning if the user executable is taking a single dash option. Prrte will parse the app arguments again later, so any changes to them here are thrown away. The executable arguments are not impacted. Refs open-mpi/ompi#10097 Signed-off-by: Austen Lauria <awlauria@us.ibm.com>

awlauria · 2022-03-11T22:17:03Z

@wzamazon I opened openpmix/prrte#1269 to try and fix the single dash issue if you want to test it.

This will preserve backwards compatability with the ompi v4 series with v5 and current master. First, this restores the silent conversion of all single dash args to multi-dash arguments. Second, it will track and eventually print all offending ompi args in a show-help message. It does this by appending all single dash arguments to an array, along with their position found in the arg list. This is key, since we don't really knoew where the user executable is on the list, and it isn't easy to tell. Luckily, pmix_arg_parse() will store where the ompi args stops in results->tail. Therefore all we need to know is if each offending argument came before results->tail or not to decide if a warning should be printed for it. This will prevent prte from erronously warning if the user executable is taking a single dash option. Prrte will parse the app arguments again later, so any changes to them here are thrown away. The executable arguments are not impacted. Refs open-mpi/ompi#10097 Signed-off-by: Austen Lauria <awlauria@us.ibm.com>

wzamazon · 2022-03-17T15:38:43Z

I can confirm openpmix/prrte#1269 fixed the issue of single dash

awlauria · 2022-03-18T15:00:55Z

This should be fixed in master with #10136. Needs back porting to v5.0.x

awlauria · 2022-03-28T13:04:15Z

~~This should be fixed in the v5.0.x branch via the following submodule pointer update: #10161~~

for main: #10165 (This PR fixes a variety of issues, but updating the submodule here will fix this issue on main).

This will preserve backwards compatability with the ompi v4 series with v5 and current master. First, this restores the silent conversion of all single dash args to multi-dash arguments. Second, it will track and eventually print all offending ompi args in a show-help message. It does this by appending all single dash arguments to an array, along with their position found in the arg list. This is key, since we don't really knoew where the user executable is on the list, and it isn't easy to tell. Luckily, pmix_arg_parse() will store where the ompi args stops in results->tail. Therefore all we need to know is if each offending argument came before results->tail or not to decide if a warning should be printed for it. This will prevent prte from erronously warning if the user executable is taking a single dash option. Prrte will parse the app arguments again later, so any changes to them here are thrown away. The executable arguments are not impacted. Refs open-mpi/ompi#10097 Signed-off-by: Austen Lauria <awlauria@us.ibm.com> (cherry picked from commit 381a521)

awlauria · 2022-03-31T21:29:23Z

Well, unfortunately I neglected to bring the fix back to the v2.1 prrte release branch. So my comment above about it being fixed in v5.0.x is incorrect (and thus is not in rc4).

This will preserve backwards compatability with the ompi v4 series with v5 and current master. First, this restores the silent conversion of all single dash args to multi-dash arguments. Second, it will track and eventually print all offending ompi args in a show-help message. It does this by appending all single dash arguments to an array, along with their position found in the arg list. This is key, since we don't really knoew where the user executable is on the list, and it isn't easy to tell. Luckily, pmix_arg_parse() will store where the ompi args stops in results->tail. Therefore all we need to know is if each offending argument came before results->tail or not to decide if a warning should be printed for it. This will prevent prte from erronously warning if the user executable is taking a single dash option. Prrte will parse the app arguments again later, so any changes to them here are thrown away. The executable arguments are not impacted. Refs open-mpi/ompi#10097 Signed-off-by: Austen Lauria <awlauria@us.ibm.com> (cherry picked from commit 381a521)

Changes: f3828e8307 - Revise show_help to use PMIx IOF 9998bfadad - schizo/ompi: Convert all single dashes to double dashes. 744710d33f - prte: check if dvm actually got set up Refs: open-mpi#10097 open-mpi#10157 Note: There are no OpenPMIx changes to pull in at this time (4/06/2022). Signed-off-by: Austen Lauria <awlauria@us.ibm.com>

awlauria · 2022-04-07T14:49:16Z

PR #10230 was merged to bring this fix to v5.0.x. This will be in rc5. Closing.

bwbarrett added Target: main Target: v5.0.x labels Mar 10, 2022

bwbarrett added this to the v5.0.0 milestone Mar 10, 2022

bwbarrett added the Severity: critical label Mar 10, 2022

awlauria mentioned this issue Mar 11, 2022

schizo/ompi: Convert all single dashes to double dashes. openpmix/prrte#1269

Merged

awlauria mentioned this issue Mar 17, 2022

Update OpenPMIx and PRRTE submodule pointers. #10136

Merged

awlauria mentioned this issue Mar 28, 2022

prterun -help fails with the error "ompi/share/prte/help-cli.txt: No such file or directory" #10168

Closed

awlauria mentioned this issue Mar 31, 2022

v2.1: schizo/ompi: Convert all single dashes to double dashes. openpmix/prrte#1322

Merged

awlauria mentioned this issue Apr 6, 2022

v5.0.x: Update PRRTE submodule pointer. #10230

Merged

awlauria closed this as completed Apr 7, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

mpirun (master branch) error message for unknown and deprecated argument is wrong #10097

mpirun (master branch) error message for unknown and deprecated argument is wrong #10097

wzamazon commented Mar 8, 2022

awlauria commented Mar 9, 2022

awlauria commented Mar 9, 2022

awlauria commented Mar 9, 2022 •

edited

Loading

wzamazon commented Mar 11, 2022

awlauria commented Mar 11, 2022

wzamazon commented Mar 11, 2022

awlauria commented Mar 11, 2022

wzamazon commented Mar 17, 2022

awlauria commented Mar 18, 2022 •

edited

Loading

awlauria commented Mar 28, 2022 •

edited

Loading

awlauria commented Mar 31, 2022 •

edited

Loading

awlauria commented Apr 7, 2022

mpirun (master branch) error message for unknown and deprecated argument is wrong #10097

mpirun (master branch) error message for unknown and deprecated argument is wrong #10097

Comments

wzamazon commented Mar 8, 2022

Background information

What version of Open MPI are you using? (e.g., v3.0.5, v4.0.2, git branch name and hash, etc.)

Describe how Open MPI was installed (e.g., from a source/distribution tarball, from a git clone, from an operating system distribution package, etc.)

If you are building/installing from a git clone, please copy-n-paste the output from git submodule status.

Please describe the system on which you are running

Details of the problem

awlauria commented Mar 9, 2022

awlauria commented Mar 9, 2022

awlauria commented Mar 9, 2022 • edited Loading

wzamazon commented Mar 11, 2022

awlauria commented Mar 11, 2022

wzamazon commented Mar 11, 2022

awlauria commented Mar 11, 2022

wzamazon commented Mar 17, 2022

awlauria commented Mar 18, 2022 • edited Loading

awlauria commented Mar 28, 2022 • edited Loading

awlauria commented Mar 31, 2022 • edited Loading

awlauria commented Apr 7, 2022

If you are building/installing from a git clone, please copy-n-paste the output from `git submodule status`.

awlauria commented Mar 9, 2022 •

edited

Loading

awlauria commented Mar 18, 2022 •

edited

Loading

awlauria commented Mar 28, 2022 •

edited

Loading

awlauria commented Mar 31, 2022 •

edited

Loading