Fix : Copy array descriptor info with associate expression#6618
Conversation
|
The change should've been very small. but how we relied on |
|
Mainly : testing CI |
|
Thanks for fixing this. Yes, let's first figure out a solution that works, then we can clean it up. This is related to #6612 also. |
- Shallow copy data ptr - Deep copy dimension descriptor + offset value This change led to : - Change on how we store `null()` in arrays (it used to set the whole descriptor to null and we fully depend on declared arrays to be ready for use)
…reateGEP` instead of relying on our map
assem2002
force-pushed
the
copy_dim_des_with_associate
branch
from
23c27e6 to
41e6b8a
Compare
…descriptor` instead of setting the whole variable to null, So we know for sure that we have the variable set with the descriptor and act upon that fact with other functions and visitors
assem2002
force-pushed
the
copy_dim_des_with_associate
branch
from
5d90cd8 to
95d9859
Compare
|
Ready to review. maybe squash merge. |
I checked locally on my machine, that this PR allows us to now run validation with POT3D with LFortran's And fixes both the MRE's reported in the linked issue. |
|
Woahh! super news. |
Pranavchiku
left a comment
Pranavchiku
left a comment
There was a problem hiding this comment.
I had a quick review, this looks good to me!
|
This is minor and can be done in subsequent PRs but let's add all the reported MREs at #6611 to integration tests. |
|
Great job @assem2002 !!! The change is bigger and I didn't review every detail, and we can review the design at our next meeting, but I think it's likely the correct fix. |
4 participants
Fixes #6611
This change led to :
null()in arrays (it used to set the whole descriptor to null and we fully depend on declared arrays to be ready for use)