Rob Latham

> To clarify, have you tried removing the whole `MUNIT_ARRAY_PARAM(argc + 1)` expression, or only just the `MUNIT_ARRAY_PARAM`? (`MUNIT_ARRAY_PARAM` is a function macro.) Lines such as ``` int munit_suite_main_custom(const MunitSuite*...

getting file size can be really expensive, but MPI_File_open is collective and we already stat the file if no file system prefix is provided. We should be able to collect...

You're off in the twisty-est, hairiest parts parts of ROMIO... We have been checking cases sort of like what you found (see https://github.com/pmodels/mpich/blob/main/test/mpi/io/resized2.c ) but I'd like very much to...

Can confirm that this patch fixes the problem for 'nfs' files. 'ufs' needs to be updated in a similar way

see https://github.com/pmodels/mpich/blob/main/src/mpi/romio/adio/common/ad_hints.c#L129 for one approach: client code sets the hint, and ROMIO will respond either with "i don't know what that is" (by not adding it to the returned info...

@hzhou this is a separate issue. We have a hint for the "synchronizing sync" behavior. This hint would allow us to further separate "visibility" from "durability" . As Adam points...

ADIO drivers are welcome to implement `ADIOI_xxx_Close` if they have a file system specific way of pushing cached items to the file system. PVFS2 for example implements an empty function

@adammoody take a peek at https://github.com/roblatham00/mpich/commit/285c7805714984c25db9700ea50bd6d6071e5d0f and if it looks good to you i'll merge to mpich main

The attached test case targets an older version of margo/mercury. I'll have to update it to our latest API

I opened this issue for the philisophical point, but in this specific case it looks like mrail requires a lot of legwork to use