Repository navigation
BiocParallel error #551
Description
Hi @cying111 ,
The error below returned when I run bambu 3.12.
My command is
se <- bambu(reads = samples.bam, annotations = annotations, genome = fa.file, ncore = 64, rcOutDir="./rcOutput/")
The error is :
--- Start generating read class files ---
Error: BiocParallel errors
1 remote errors, element index: 1
0 unevaluated and other errors
first remote error:
Error in .Call2("C_find_overlaps_in_groups_NCList", start(q), end(q), : negative length vectors are not allowed
Execution halted
Do you have an idea how to troubleshoot? I ran on a different sample without problem.
Thanks,
Sophia
Activity
sophiachen1 commentedon Apr 9, 2026
I solved this issue by adding lowMemory option.
sophiachen1 commentedon Apr 16, 2026
Hi team,
I am sorry to reopen this again. The same error comes back for some other samples even with the lowMemory option on. Do you have any suggestion to troubleshoot it?
Thank you so much,
Sophia
sophiachen1 commentedon Apr 20, 2026
Hi @cying111,
Have you seen this error before? I think it happens to my bam files more than 50 Gb. Any idea to troubleshoot?
Thank you ahead,
Sophia
yinshiyi commentedon Oct 12, 2026
@sophiachen1 ,
I think it is not a pure negative error, it possibly indicate that your sample is rich.
Possibly root cause: in a deep dataset, thousands of distinct high-confidence spliced read classes appears, findOverlaps() call returns more than 2³¹ (about 2.1 billion) hits, and IRanges can’t store that many.
In C behavior, hit count goes past INT_MAX, the length wraps to a negative number. R’s allocVector then fails with “negative length vectors are not allowed”.
i think looking into the design of
getUnsplicedReadClassByReferenceto improve the split worker design.in lowMemory, the work is splits by chromosome. that fixes samples where the total classes across the genome went over 2³¹. but when the dataset got even bigger, split by chromosome will still crash.