doc: Add links to disable-io-romio option #11468
Conversation
|
I guess the real question is: why do we (still) have If it's effectively an alias for Also: what happens if you specify |
wckzhang
force-pushed
the
disable_romio_doc
branch
from
404b954 to
9d73da9
Compare
|
I'm not sure about the history but when I was reading the code it seemed to just be an enable/disable feature for the romio io component (./ompi/mca/io/romio341/configure.m4) though I don't know how standard this way of disabling component compilation is. As far as I can tell it's just used in the romio component and it doesn't look to disable MPI-IO. From a functional standpoint it basically looks to do the same thing as --enable-mca-no-build. From the code itself it looks like it'd be able to accept both --disable and no build as the disable looks to be localized to the romio component. I can add a comment that it functionally works as enable-mca-no-build for the romio341 component if you'd like |
Yeah, that would probably be good. We might even want to deprecate this option, since a) it is doing something different than it used to, and b) the new functionality that it does has a more general mechanism that accomplishes the same thing. |
Do we have a standard procedure for deprecating options? I don't want to pull the rug out from somebody relying on this build flag. |
We generally try to emit a warning for at least one major release series (i.e., 5.x), and then talk about removing it in the next major release series. I.e., if someone executes |
wckzhang
force-pushed
the
disable_romio_doc
branch
from
9d73da9 to
be2a6e5
Compare
| from being compiled.This is an outdated mechanism and functions | ||
| similarly to ``--enable-mca-no-build`` for the romio341 component. |
There was a problem hiding this comment.
| from being compiled.This is an outdated mechanism and functions | |
| similarly to ``--enable-mca-no-build`` for the romio341 component. | |
| from being compiled. This is a deprecated mechanism; it is effectively the | |
| same as adding ``io-romio341`` to the list of items passed to ``--enable-mca-no-build``. |
There was a problem hiding this comment.
Does anyone else find it disturbing that the user has to know the version of ROMIO embedded in the OMPI code base just so they can disable it?
There was a problem hiding this comment.
I think the problem is more that the romio component was poorly named
There was a problem hiding this comment.
I think the problem is more that the romio component was poorly named
I think you're right; when this component was initially created, we hadn't envisioned anything other than ROMIO, and we had envisioned being able to support more than one version of ROMIO simultaneously.
Neither of those have really turned out that way (i.e., we emphasize OMPIO these days, and I think we've only ever had multiple versions of ROMIO for developers -- not end users). I wonder if we should take a future action item to rename this component romio and add an MCA component name synonym in for romio341 (for backwards compatibility -- similar to what we did for vader).
Added links to other documents explaining ROMIO and describe it being a deprecated option similar to --enable-mca-no-build Signed-off-by: William Zhang <wilzhang@amazon.com>
wckzhang
force-pushed
the
disable_romio_doc
branch
from
be2a6e5 to
97b4938
Compare
| @@ -68,6 +68,8 @@ The following *projects* exist in Open MPI |ompi_ver|: | |||
| See :ref:`the role of PMIx and PRRTE | |||
| <label-running-role-of-pmix-and-prte>` for more information. | |||
|
|
|||
| .. _label-mca-frameworks: | |||
There was a problem hiding this comment.
You may need to rebase / fix conflicts with these changes after merging #11459.
There was a problem hiding this comment.
Aye, I'll go fix it before merging
|
Yeah I can do that |
|
Backport - #11554 |
Added links to other documents explaining ROMIO