fix: the graph calculation with post_depends.#4539
Conversation
|
It fixes: |
|
looks identical to #4404 |
|
Not sure why this isn't correct? I.e. the dependency is 'inversed'. |
|
Oke so the issue is that this will result in So when you run two and three, it would run one even tough it isn't in the running tasks. So what we actually want is: but when you have the dependencies of two you want to get: So maybe we can correct this by looking at the post deps first, if there is a post_dependency |
|
I suspect the post_dependency is really a feature that increases the changes by a lot that you get 'cyclical' graphs. Tracing the code: The tasks vector we get to |
|
I've run the deps command locally and that one is still incorrect: So i suppose the deps needs to be updated that it shows |
|
@jdx Is there any way we can move this forward, it fixes the post_depends lockup bug. |
|
I hope you can find some time to have another look at this to atleast have a better working post_dependency instead of a 'fully broken' version of it. |
This bugfix fixes the graph calculation it seems that the 'post_depends' where incorrectly handled resulting in tasks where all dependending on eachother: