Repository navigation
Race condition in the shared memory implementation, missing sync in read_sm and write_sm #195
Description
Currently, In the shared memory c++ implementation If the writer updates the simulation state while the reader is reading it, the reader might get corrupted data (half old state, half new state).
The write_SM function writes to the shared memory buffer using without acquiring a lock, which will let the read_sm function 'read' simultaneously when write_Sm is running 'write'. This should not happen, this will result in corrupted data to the reader.
Metadata
Metadata
Assignees
Labels
No labels
Activity
GREENRAT-K405 commentedon Feb 2, 2026
@pradeeban I would be interested in solving this issue. I had just learned something of similar concept(locking) in university the previous semester (in DBMS course). Would love to practically implement the concept.
GREENRAT-K405 commentedon Feb 3, 2026
@pradeeban do you think we should have some more samples in demo especially for languages other than python? If so do let me know I will create an separate issue for that.
I will be signing off for today, will try to get as many issues resolved by tomorrow!
pradeeban commentedon Feb 4, 2026
@GREENRAT-K405 More samples are always nice. But as of now, without the real use case running on, these will likely become toy samples, which I do not like much. Let's see.
fix(shm): serialise concurrent writers with seq# + POSIX semaphore (f…
fix(shm): serialise concurrent writers with seq# + POSIX semaphore (f…
fix(shm): serialise concurrent writers with seq# + POSIX semaphore (f…
fix(shm): serialise concurrent writers with seq# + POSIX semaphore (f…
pradeeban commentedon Jun 30, 2026
Thanks, merged.