Check SciFi banks and sort them by hand
This will slow down the decoding quite a bit but we check for scenarios that should never happen and yet do in the data.
- if a bank is sent with no header (would create a segfault);
- if a bank does not have an even size;
- if a bank has an absurd number of clusters (> number of links * maximum number of clusters per link);
- if a link is larger than 24 (would create a segfault);
- if a link sent more than 24 clusters.
Additionally, we sort clusters directly. Unsorted clusters should only show up in station 3 but seem to rarely occur somewhere else. An issue will be created.
We keep small_vectors rather than static_vectors to avoid a segfault in case a link sent more than 24 clusters.
Each of these counters now comes with an histogram which helps locate the bank or unique link that was responsible.
Merge request reports
Activity
assigned to @lohenry
added RTA label
mentioned in issue #247 (closed)
mentioned in issue Moore#443 (closed)
mentioned in merge request !3720 (closed)
mentioned in merge request !3733 (closed)
added 1 commit
- ce6e95e1 - Change the NbBanks number to NbBanksMax. Correct findBankAndLink. Add odd bank...
added 1 commit
- 63b9684b - Make a difference between types of corrupt clusters and add histograms
added 1 commit
- 769935da - Adding an error counter if cluster vectors get too full
- Resolved by Rosen Matev
/ci-test
added ci-test-triggered label
- [2022-08-20 16:30] Validation started with lhcb-master-mr#5460
- Resolved by Louis Henry
@lohenry Is this ready for testing and merging ? If so please can you fix the pipeline formatting errors ?