Data.table: fcase() may trigger segfaults in certain cases

[fredguinog]

I'm using data.table 1.12.9, the most recent development version

Minimal reproducible example

qtty_row_table <- 1000000L

a <- data.table::data.table(
  V1 = rnorm(qtty_row_table),
  V2 = rnorm(qtty_row_table),
  V3 = rnorm(qtty_row_table)
)

a[, v0 := data.table::fcase(
  V1 > 0 & V2 <= 1 & V3 > 1, V2 * 100L,
  V1 > 1 & V2 <= 0 & V3 > 0, V3 * 100L,
  V1 > -1 & V2 <= 2 & V3 > 1, V1 * 100L,
  V1 > 1 & V2 <= 0 & V3 > 2, 300,
  V1 > 0 & V2 <= 1 & V3 > 1, 100,
  V1 > -1 & V2 <= 0 & V3 > -1, V1 * 100L,
  default = 0
)]

Output of sessionInfo()

R version 3.5.2 (2018-12-20)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 18363)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252  LC_CTYPE=English_United States.1252    LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                           LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] compiler_3.5.2 tools_3.5.2   

[MichaelChirico]

Working fine for me on Mac / 3.6. Could you share more about the crash? Is it a segfault? Out of memory?

R version 3.6.0 (2019-04-26)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS Mojave 10.14.6

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib

locale:
[1] C/UTF-8/C/C/C/C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] compiler_3.6.0    data.table_1.12.9

[fredguinog]

The rsession just crashes without showing any message. Using the event viewer from windows 10, it generates this log:

Log Name:      Application
Source:        Application Error
Date:          4/13/2020 11:00:49 PM
Event ID:      1000
Task Category: (100)
Level:         Error
Keywords:      Classic
User:          N/A
Computer:      LAPTOP-RLEB1RA2
Description:
Faulting application name: Rterm.exe, version: 3.52.10334.0, time stamp: 0x5c1b62e0
Faulting module name: R.dll, version: 3.52.10334.0, time stamp: 0x5c1b62d0
Exception code: 0xc0000005
Fault offset: 0x0000000000051373
Faulting process id: 0x4f84
Faulting application start time: 0x01d6120081d45d88
Faulting application path: C:\PROGRA~1\R\R-35~1.2\bin\x64\Rterm.exe
Faulting module path: C:\PROGRA~1\R\R-35~1.2\bin\x64\R.dll
Report Id: fce4b052-d417-4390-ac37-5e28bfc2d145
Faulting package full name: 
Faulting package-relative application ID: 

[shrektan]

I can reproduce this on Windows R3.5.3.

---

UPDATE: Super weird. Can't reproduce on Windows R3.6.2 so I believe it only occurs on Windows with R versions older than or equal to R3.5.3.

On R3.6.2 it crashes as well but needs to run it without RStudio (in RStudio you have to run it repeatedly for 2-3 times before it crashes).

[MichaelChirico]

Also works for me on rocker/r-ver:3.5.3 so it must be a windows thing

[jangorecki]

Unless there will be a report of this issue on windows R > 3.5.3 I think it is safe close.

[ChristK]

I can reproduce in R 3.6.2 on W10. It crashes R every time (not using Rstudio). In Rstudio sometimes I have to run the code 2 or 3 times before it crashes. But it always crashes eventually.

[MichaelChirico]

this build of fcase-windows branch also failed on appveyor

https://ci.appveyor.com/project/Rdatatable/data-table/builds/32166226

[shrektan]

@ChristK You are right. It crashes in R3.6.2 on Windows. I didn't see it because I run it in RStudio. In the terminal, the R session just quits (assumes crashed).

[sindribaldur]

Crashed on first run in Rstudio, R 4.0.0, W10

[shrektan]

According to my test, it should have been fixed by #4401. However, additional verifications are welcome (install the PR via remotes::install('rdatatable/data.table#4401') )

[fredguinog]

It worked like a charm
Thank you very much

[shrektan]

Thanks for the response but we need to reopen it as it should be closed after the PR gets merged to the master branch.

[2005m]

@fredguinog , you can use kit::nif (look at the news file for more info) until fcase gets fixed. You might be waiting for some times. You might also be interested by kit::vswitch which is an order of magnitude faster.

[mattdowle]

Thanks for the example and details. Turns out it was fixed by #4629 before fcase was released in 1.13.0. Thanks to static code analysis by the rchk tool in release procedures. Point 2 mentions rework to fifelse.c and the first value not being protected.
Sorry I didn't see your good example, or @shrektan's PR #4401 at the time. Double-checks to be sure in https://github.com/Rdatatable/data.table/pull/4401#issuecomment-720944605.