Setting alloccol and then making a DT with more than that many columns segfaults

[caneff]

I have datatable.alloccol set in my options (not even sure why any more, probably from helping with a rare segfault debug in an earlier version). When I make a DT with more than that many columns it segfaults. With the latest dev version and R 3.0.2:

options(datatable.verbose=TRUE)
options(datatable.alloccol=100)
library(data.table)
a=data.frame(matrix(1,ncol=100))  
data.table(a) # Works fine
a=data.frame(matrix(1,ncol=101))  
data.table(a) # Segfaults

The output when it segfaults (with things in {} removed):

> data.table(a)
*** glibc detected *** {path to R}/R: malloc(): memory corruption: {memory address} ***

I can just remove alloccol for now as I'm not sure it is still necessary. No alloccol and it works fine.

[arunsrinivasan]

Please read completely and follow the instructions here. Thank you.

[caneff]

I don't understand what I'm missing. I told you it is the latest dev version of data.table and R 3.0.2, I gave a minimal reproducible example, and I don't see this bug already reported. What more are you looking for? Okay fine, it is isn't totally minimal as I was trying to show you where the edge case issue clearly is. The real minimal example:

options(datatable.alloccol=100)
library(data.table)
a=data.frame(matrix(1,ncol=101))  
data.table(a) 

[arunsrinivasan]

@caneff
When I mean "completely", I really mean it. Even the > in code, and # for comments. On a second look, in this case, it seems to not matter much, but then it makes it clear that you've completely read the instructions.

It's also easier for us to if one shows the output (wherever relevant) - which is not necessary in your case again, but I wrote the comment here so that it doesn't happen on another post.

Thanks again. I don't get a segfault on my machine, but it errors instead:

# Error in alloc.col(ans) : attempt to set index 100/100 in SET_VECTOR_ELT
# data.table 1.9.5, R v3.1.2, Mac OS X, Yosemite 10.10.1

[arunsrinivasan]

@caneff, have patched it. I don't see the error on my machine anymore. Could you please confirm on segfault please? Thanks.