[PATCH] spec/bsps: Deduplicate objxilinxsupport

[Chris Johns]

On 7/2/2023 5:55 pm, Sebastian Huber wrote:
> On 06.02.23 23:09, Gedare Bloom wrote:
>> ok. I'm not sure, maybe we need a note about this in
>> https://docs.rtems.org/branches/master/eng/build-system.html
> 
> A hint in the documentation would be helpful. Even more helpful would be a
> consistency check before patch sets are pulled in. For example, it is a bug if
> you have a diamond in the build dependency graph and the nodes with an indegree
> greater than one install files. If the nodes with an indegree greater than one
> build source files with identical build options, then this is also a bug.
> However, a valid special case could be a node which is built with different
> build flags, for example:
> 
> lib -> grp_a (-DFLAG_A) -> objxyz
>  \---> grp_b (-DFLAG_B) -> objxyz

I would expect that case would map to separate build objects?

How do you suggest someone proceed to add a check?

Would that check happen when the spec files are parsed? That is, we assume the
pickled output loaded on each build is OK?

Do we have any unittests for the build system? Adding a check would mean we need
a test case to generate the problem?

Chris