Would it make sense to have a XOR condition ?

[fred-r]

Today with the conditions we have 3 possibilities:

accept1	At least one accept must be true to signal a true for the complete condition (OR-Rule)	FilterType	0..*
require1	All require must be true to signal a true for the complete condition (AND-Rule).	FilterType	0..*
deny1	If one deny is true the complete condition becomes false. This element overrules require and accept (AND-NOT-Rule).	FilterType	0..*

ACCEPT is an OR so both options can be selected at the same time.
Would it be useful to have an XOR: N possibilities but only 1 can be selected.

[fred-r]

@cparata

[edriouk]

@fred-r, we have discussed that in our team in the past. We found that useful to have a "one of" expression.
However have not decided on how to implement it in a way that supports backward and forward compatibility.
The major problem is adding a new tag. Older tools will simply ignore those elements.
Adding an attribute to an accept expression won't work either, because:

• it would not pass validation if an attribute does not match expression domain
• need to be specified to all accept expressions, otherwise can cause an inconsistency

Another option is to add an "exclusive" attribute to a condition. This will be also ignored by older tools, but I think it could be tolerated.

Do you have any other suggestion how to implement it?

Please also note that currently we have three options for exclusivity that can be used instead of condition:

1. Cbundle: works on Cclass level
2. Exclsusive API: works on Cgroup level
3. Cvariant: works on Cgroup and Csub levels.

[jkrech]

@edriouk is this what you have in mind?

<condition id="Exclusive RTOS" exclusive="true">
  <accept Cclass="CMSIS" Cgroup="RTOS2" Csub="RTX5" />
  <accept Cclass="CMSIS" Cgroup="RTOS2" Csub="FreeRTOS" />
</condition>

If "exclusive" attribute is set, then the body must only contain <accept> elements describing component attributes (Cvendor, Cbundle, Cclass, Cgroup, Csub, Cvariant, Cversion).

[edriouk]

@jkrech , yes that is my suggestion.

[mdortel-stm]

It looks a nice way to bring the feature.

[ReinhardKeil]

Is this not duplicating a feature that already exists with exclusive=1 at level.
https://github.com/ARM-software/CMSIS_6/blob/main/ARM.CMSIS.pdsc#L445

I believe this does effectively the same:

<condition id="RTOS2">
  <accept Cclass="CMSIS" Cgroup="RTOS2"  />
</condition>

[mdortel-stm]

@ReinhardKeil the exclusive attribute you refer to applies only on APIs. IMO, it does not match the original topic.