You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Chunks of docs text in src/docs/prrte-rst-content. This content is generally included and/or embedded in other docs (such as the docs in 1, but also the Open MPI mpirun(1) man page).
"Show help" text files in src/docs/show-help-files. This content is specifically formatted displaying error messages at run time (vs. being pretty documentation that is rendered into HTML), but also includes a bunch of content from 2.
A standalone file in src/mca/schizo/ompi/schizo-ompi-cli.rstxt that is specifically intended for Open MPI's mpirun(1) command documentation (man page). It also includes a bunch of content from 2.
Open MPI's docs (in its docs/ directory) copies PRTE's src/docs/prte-rst-content and the schizo-ompi-cli.rstxt to OMPI's docs/prte-rst-content/ so that it can build + render the overall OMPI docs.
Task
After moving PRTE out of a submodule and into Open MPI proper, remove the logic in Open MPI's docs/Makefile.am that copies over the prrte-rst-content RST files from the internal or external PRTE (remember: now, there will never be an external PRTE) to OMPI's docs/ dir. Instead, replace this logic with a relative sym link directly to the location of the PRTES RST source files in the OMPI git clone.
Also move the schizo ompi CLI rst file into the target location in the docs tree rather than having it live in the schizo ompi component (since the OMPI PRTE is now OMPI-specific, there is no need for the schizo framework, and we can be OMPI-specific).
Since we will always be building Open MPI against the internal PRTE, remove logic (and any additional scripts or data files) that handles the case where we were not building against any PRTE.
NOTE: This issue specifically does not cover consolidating the main PRTE docs in 1 (above). That will be a separate issue.
The text was updated successfully, but these errors were encountered:
since the OMPI PRTE is now OMPI-specific, there is no need for the schizo framework
Pointed this out elsewhere, but might as well correct this statement here too. You must retain the schizo framework as the prted daemons require the prte personality. Removing it also means that fixes/changes made upstream are suddenly no longer portable - and that probably isn't a desirable situation.
So can it be removed? Sure, with a lot of work and plenty of risk. Might be worth some careful thought about how far to go in this direction.
Background
There are 4 separate sets of PRTE docs (some of which get combined):
docs/
(as you can see in https://docs.prrte.org/)src/docs/prrte-rst-content
. This content is generally included and/or embedded in other docs (such as the docs in 1, but also the Open MPImpirun(1)
man page).src/docs/show-help-files
. This content is specifically formatted displaying error messages at run time (vs. being pretty documentation that is rendered into HTML), but also includes a bunch of content from 2.src/mca/schizo/ompi/schizo-ompi-cli.rstxt
that is specifically intended for Open MPI'smpirun(1)
command documentation (man page). It also includes a bunch of content from 2.Open MPI's docs (in its
docs/
directory) copies PRTE'ssrc/docs/prte-rst-content
and theschizo-ompi-cli.rstxt
to OMPI'sdocs/prte-rst-content/
so that it can build + render the overall OMPI docs.Task
After moving PRTE out of a submodule and into Open MPI proper, remove the logic in Open MPI's
docs/Makefile.am
that copies over theprrte-rst-content
RST files from the internal or external PRTE (remember: now, there will never be an external PRTE) to OMPI'sdocs/
dir. Instead, replace this logic with a relative sym link directly to the location of the PRTES RST source files in the OMPI git clone.Also move the schizo ompi CLI rst file into the target location in the docs tree rather than having it live in the schizo ompi component (since the OMPI PRTE is now OMPI-specific, there is no need for the schizo framework, and we can be OMPI-specific).
Since we will always be building Open MPI against the internal PRTE, remove logic (and any additional scripts or data files) that handles the case where we were not building against any PRTE.
NOTE: This issue specifically does not cover consolidating the main PRTE docs in 1 (above). That will be a separate issue.
The text was updated successfully, but these errors were encountered: