Closed mblanche closed 3 years ago
I can confirm that I can reproduce the problem. This seems sufficient to crash R:
library(InteractionSet)
library(rhdf5)
h5read("toy1.mcool", name = "/resolutions/1000/bins")
# corrupted double-linked list
# [1] 4414 abort (core dumped) R
Oddly, this seems to work without issue
h5read("toy1.mcool", name = "/resolutions/2000/bins")
And then its possible to read the 1000
group immediately afterwards without a crash. I'll keep digging.
I think I've tracked this down. Can you give this version a try?
BiocManager::install('grimbough/rhdf5', ref = 'RELEASE_3_11')
For me it can now run both parts of your example code without issue, fingers crossed it the same for you.
Ok, seems like it's working. Thanks a lot for the fast answer. I'll let you know if anything else pops up
Great, thanks for reporting back. I'll push the changes to Bioconductor, should take a couple of days before it's available via that route.
Hi, for the past two weeks, I've been experciencing segmentation fault with rhdf5. Initially, everything was working fine but I think that I got a new version that started crashing.
The segmentation fault seems to be OS specific. It's happening on Ubuntu 18.04.5 LTS, using R 4.02 but not on Mac. For some reason running sessionInfo() after loading rhdf5 give me two versions, is this normal?
Also, I have a toy case that shows this issue. I have a function, getBins() reading some data from a toy rhdf5 file (https://dovetail-public.s3-us-west-2.amazonaws.com/rhdf5/toy1.mcool). When on Ubuntu, loading rhdf5 prior to InteractionSet let me pull the data from the file, while if I load InteractionSet before rhdf5, I get a segmentation fault every time. One of my colleague tested on a Mac OS X Catalina with no issues but was able reproduce the problem on Ubuntu.
Here's the script, the file toy rhdf5 file can be dowloaded here https://dovetail-public.s3-us-west-2.amazonaws.com/rhdf5/toy1.mcool
Any chance you guys could look at this?
Many thanks
Marco