"Index requested greater than vector's size" when loading pufferfish index using salmon v1.0

[mdshw5]

Is the bug primarily related to salmon (bulk mode) or alevin (single-cell mode)?
salmon
Describe the bug
When loading a pufferfish index using salmon v1.0 I encounter errors. I've posted to COMBINE-lab/pufferfish#8 with a full description.

When running salmon v1.0 using a rather large index, I receive an error raised from the cereal library "Index requested greater than vector's size". The log reads:

-----------------------------------------
| Loading contig table | Time = 12.954 s
-----------------------------------------
size = 35010142
-----------------------------------------
| Loading contig offsets | Time = 269.18 ms
-----------------------------------------
-----------------------------------------
| Loading reference lengths | Time = 7.8427 ms
-----------------------------------------
-----------------------------------------
| Loading eq table | Time = 3.3896 s
-----------------------------------------
-----------------------------------------
| Loading mphf table | Time = 3.8301 s
-----------------------------------------
size = 3567796961
Number of ones: 35010141
Number of ones per inventory item: 512
Inventory entries filled: 68380
-----------------------------------------
| Loading contig boundaries | Time = 11.288 s
-----------------------------------------
size = 3567796961
-----------------------------------------
| Loading sequence | Time = 7.763 s
-----------------------------------------
size = 2517492731
-----------------------------------------
| Loading positions | Time = 171.81 s
-----------------------------------------
size = 3221360466
-----------------------------------------
| Loading reference sequence | Time = 7.9564 s
-----------------------------------------
-----------------------------------------
| Loading reference accumulative lengths | Time = 35.741 ms
-----------------------------------------
Index requested greater than vector's size: 6442720932>6442720932
Index requested greater than vector's size: 6442720996>6442720932
Index requested greater than vector's size: 6442721060>6442720932
Index requested greater than vector's size: 6442721124>6442720932
Index requested greater than vector's size: 6442721188>6442720932
Index requested greater than vector's size: 6442721252>6442720932
Index requested greater than vector's size: 6442721316>6442720932
Index requested greater than vector's size: 6442721380>6442720932
Index requested greater than vector's size: 6442721444>6442720932
...

The index does not finish loading, and so salmon does not enter read quantification routines.

To Reproduce

• Which version of salmon was used? 1.0
• How was salmon installed (compiled, downloaded executable, through bioconda)?
  Github release tarball
• Which reference (e.g. transcriptome) was used?
  Gencode v32, with additional elements representing genic introns and intergenic spaces.
• Which read files were used?
  NCBI SRA run accession GSM2392582
• Which which program options were used?
  --no-version-check --libType ISR --threads 4 --seqBias --gcBias --useVBOpt

Expected behavior
I expected index loading to complete successfully.
Desktop (please complete the following information):

• OS: CentOS6

Additional context
I've uploaded the index file archive here and the transcripts fasta file here.

[mdshw5]

Sorry, I submitted this prematurely - let me edit the description.

[mdshw5]

I'm cross-posting this issue to COMBINE-lab/pufferfish#8, so please feel free to close one or the other. Also, I just noticed the archives of index files and transcripts I uploaded are named "rapmap" - that was definitely just a mental slip and I meant "pufferfish" :).

[mdshw5]

Closing for now as the discussion occurred on the pufferfish repo.