readChangeoDb failed in sbatch mode on Farnam

[Julian Zhou]

 *** caught illegal operation ***
address 0x2ac521618728, cause 'illegal operand'

Traceback:
 1: .Call("readr_guess_header_", PACKAGE = "readr", sourceSpec, tokenizerSpec,     locale_)
 2: guess_header_(datasource, tokenizer, locale)
 3: guess_header(ds_header, tokenizer, locale)
 4: col_spec_standardise(data, skip = skip, comment = comment, guess_max = guess_max,     col_names = col_names, col_types = col_types, tokenizer = tokenizer,     locale = locale)
 5: read_delimited(file, tokenizer, col_names = col_names, col_types = col_types,     locale = locale, skip = skip, comment = comment, n_max = n_max,     guess_max = guess_max, progress = progress)
 6: readr::read_tsv(file, na = c("", "NA", "None"))
 7: withCallingHandlers(expr, message = function(c) invokeRestart("muffleMessage"))
 8: suppressMessages(readr::read_tsv(file, na = c("", "NA", "None")))
 9: readChangeoDb(paste0("d", d, "_14_cat.tab"))
An irrecoverable exception occurred. R is aborting now ...
/var/spool/slurmd/job1509736/slurm_script: line 14: 39668 Illegal instruction

Solution: use base R

read.table("changeoDb.tab", sep="\t", header=T, stringsAsFactors=F)

[Jason Vander Heiden]

So... We can't switch to base R for performance reasons (see #39). Hrm. I wonder if this is GCC related?

[Julian Zhou]

Welp, remember that one time we tried numerous things including changing GCC versions on Farnam and nothing worked? That was for mutate_at/each from dplyr, but I suspect the same thing won't for readr either..

[Jason Vander Heiden]

Yeah, I'm sure you're right. It stands to reason we'll have this problem with dplyr in general using sbatch.

[Julian Zhou]

Ergh last time I was able to work around to the problem by running interactive instead of sbatch, but now it fails in interactive mode too :/ #freakingout

[Jason Vander Heiden]

No bueno. I'll test tomorrow. See if I can figure it out.

[Jason Vander Heiden]

Apparently, we need to compile tidyverse packages on the m915 node or they won't work on the m915 node. See the Problematic R package compilation section of the cluster wiki:
https://bitbucket.org/kleinstein/immcantation/wiki/ClusterConfig