Fails torture #2841
Milestone
Comments
|
In particular, > library("data.table", lib.loc = tempd)
data.table 1.11.1 IN DEVELOPMENT built 2018-05-05 01:21:17 UTC; travis
The fastest way to learn (by data.table authors): https://www.datacamp.com/courses/data-analysis-the-data-table-way
Documentation: ?data.table, example(data.table) and browseVignettes("data.table")
Release notes, videos and slides: http://r-datatable.com
> dt = data.table(a = rep(as.Date("2010-01-01"), 4), b = rep("a",4))
> identical(rbind(dt,dt), data.table(a = rep(as.Date("2010-01-01"), 8), b = rep("a",8)))
[1] TRUE
> gctorture2(step = 1)
[1] 0
> dt = data.table(a = rep(as.Date("2010-01-01"), 4), b = rep("a",4))
> identical(rbind(dt,dt), data.table(a = rep(as.Date("2010-01-01"), 8), b = rep("a",8)))
[1] FALSE
> rbind(dt, dt)
b a
1: a 2010-01-01
2: a 2010-01-01
3: a 2010-01-01
4: a 2010-01-01
5: a 2010-01-01
6: a 2010-01-01
7: a 2010-01-01
8: a 2010-01-01 |
Merged
|
Great. That |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
(Writing with future reference in mind in case we hit this again.)
Motivated by #2767, we're looking for a memory fault in data.table. At least, we have to utterly rule out everything on the data.table side, first. Experience shows that R-devel-only problems are in fact revealing underlying problems in data.table. Particularly on 32bit which seems to expose memory problems faster, or when R changes internal memory layout, as happened recently in R 3.5.0.
All rchk messages have been dealt with in #2836. All were false positives I believe. Still, it was useful to flag, consider and make rchk-compliant so we don't have to consider those lines again.
rchkis currently set to be 'permissive' according to notes on CRAN and its documentation. Perhaps now we pass 'permissive' (when 1.11.2 goes to CRAN), we can ask for a stricterrchksetting for data.table, or run it ourselves. We could come back to that.However, with latest R-devel, compiled with ASAN and
--strict-barrierandtorture2(step=100), we fail with the following log. The runtime was somewhere between 12h and 18h. (It was still running when I checked at 12h, and had failed when I checked again at 18h).Log
The text was updated successfully, but these errors were encountered: