Why does PMI_Data_Pack description include "These errors cannot be detected during packing"

[dsolt]

Describe your question or clarification request here

In the Packing/Unpacking chapter there is text such as:

Note that any data to be packed that is not hard type cast (i.e., not type cast to a specific size) may lose precision when unpacked by a non-homogeneous recipient. The PMIx_Data_pack function will do its best to deal with heterogeneity issues between the packer and unpacker in such cases. Sending a number larger than can be handled by the recipient will return an error code (generated upon unpacking) — the error cannot be detected during packing.

If we know the process we are packing for, then presumably PMIx could gather whatever information it needed about the destination to be able to generate this error on the packing side. Why does it claim the error cannot be detected during packing?

Suggested Clarification

1. explain why this is true if it is
2. Remove it
3. Say that "Implementations may choose not to detect this until it is unpacked"

References

Please list specific chapters/sections in the PMIx Standard related to your query.

[abouteiller]

Discussion during ASC24Q2

• the target provides the pmix version info, but it may not contain the target architecture (Scalability reasons)
• hence the packing may be target agnostic: can't see the error at that point

TODO: decide if we need more clarification, or close issue altogether.

[abouteiller]

Just change the error cannot be detected during packing. to may not, the rest is clear enough