Hi,
I am tracing down a problem with qvalue 1.39.1 (current version from BioC 3.0). I have been running some batch jobs that take 3-5 days, get a set of p-values and then use qvalue::qvalue(). Some of them are working ok, but two recently (and a third from another user) segfaulted. It could be an issue with our cluster, or maybe there is a problem with the input p-values (I didn't save them so I have to re-generate them to confirm this). Anyhow, I am just curious if others have experienced problems like this.
Error output on on *.e* file
*** caught segfault ***
address (nil), cause 'unknown'
Traceback:
1: smooth.spline(lambda, pi0, df = smooth.df)
2: qvalue(regs$pvalues)
## note regs$pvalues is the object that is created prior to using qvalue::qvalue() and has the p-values
## There's also no other info, for example, if qvalue::qvalue() detected a problem with the input.
As proof that this doesn't always happen, I could use the example data without any problems.
> library(qvalue)
> data(hedenfalk)
> q <- qvalue(hedenfalk)
> head(q$qvalues)
[1] 0.0873481 0.2073648 0.6616095 0.1600905 0.6264245 0.3681680
> devtools::session_info()
Session info-------------------------------------------------------------------
setting value
version R version 3.1.1 Patched (2014-10-14 r66758)
system x86_64, linux-gnu
ui X11
language (EN)
collate en_US.UTF-8
tz <NA>
Packages-----------------------------------------------------------------------
package * version date source
colorout * 1.0.2 2014-05-30 local
devtools 1.6.1 2014-10-07 CRAN (R 3.1.1)
qvalue * 1.39.1 2014-05-30 Bioconductor
rstudioapi 0.1 2014-03-27 CRAN (R 3.1.1)
Our other lead is that in our cluster, R 3.1.x gets re-compiled every morning and in some cases this could break things. It's the price payed for keeping everything updated (once BioC 3.0 is released, stability over up-datedness will be favored).
Thanks,
Leo
It'll be helpful to provide an example of the data that *does* segfault, e.g., a subset of the vector
regs$pvalues; as you say the example provided does not, so hard to track down...I am in the process of re-generating that vector. But for now we are leaning towards this being caused by our cluster setup.
I too am having this problem. Did you figure out a solution?
Sorry, this was so long ago that I don't remember. I even forgot about this thread and as you can see, I never got around to providing a reproducible example. Guess I can use this thread as an example of what not to do in the Bioc support website.
My guess from using `qvalue` in other projects is that it sometimes fails if nearly all the p-values are very low to begin with (like no data from 0.5 to 1). Anyhow, if you are having this same problem I recommend that you post a new question with a reproducible example so that you'll get an answer.
Best,
Leonardo