Debian testing failure for gprod test about data.table

[ben-schwen]

@MichaelChirico Should we close since duplicate to #6137?

[eddelbuettel]

Sorry, I wish I could help but I have particular insight here. My world is amd64. Debian has porter teams. Maybe the Debian maintainers for data.table can help you here and contact them -- I was just trying to unplug this situation as the data.table thread of removal also threatens 10+ packages I look after.

[MichaelChirico]

cc @ben-schwen.

From the failure log, it looks like the testing environment has a strange setup?

sizeof(long double)==8,
longdouble.digits==, sizeof(pointer)==4, TZ==unset

longdouble.digits is empty?

data.table/R/test.data.table.R

Lines 230 to 231 in cbe491f

	", sizeof(long double)==", .Machine$sizeof.longdouble,
	", longdouble.digits==", .Machine$longdouble.digits, # 64 normally, 53 for example under valgrind where some high accuracy tests need turning off, #4639

For me that's

.Machine$longdouble.digits
# [1] 64

That test is behind if (test_longdouble) which in turn examines capabilities("long.double"). Any idea what might be the reason @eddelbuettel?

[MichaelChirico]

From how .Machine is populated, it's strange to see nothing in longdouble.digits:

https://github.com/r-devel/r-svn/blob/5c0e367225ec0ed9e1a95864302a6311708533ee/src/main/machine.c#L186

That #ifdef HAVE_LONG_DOUBLE branch is the same as populates .Machine$sizeof.longdouble. So there should be some integer there. I don't follow what this routine is doing to define that it value that winds up as longdouble.digits:

https://github.com/r-devel/r-svn/blob/5c0e367225ec0ed9e1a95864302a6311708533ee/src/main/machar.c#L95

Anyway, time to follow up with the maintainers.

[eddelbuettel]

For some (but not all?) architectures we can get shell access, the table is here and the first one on the list armhf. So if guys have something testable I can probably find a moment to run a quick test.

[MichaelChirico]

Hmm just reporting back .Machine and capabilities() on that system would be helpful.

As a workaround we can probably just make test_longdouble also look at length(.Machine$longdouble.digits) > 0L also.

[eddelbuettel]

Here you go:

(sid_armhf-dchroot)edd@abel:~$ R
                                                    
R version 4.4.0 (2024-04-24) -- "Puppy Cup"
Copyright (C) 2024 The R Foundation for Statistical Computing
Platform: arm-unknown-linux-gnueabihf (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
                                                    
R is a collaborative project with many contributors. 
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
                                                    
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> .Machine        
$double.eps
[1] 2.220446e-16
                                                    
$double.neg.eps
[1] 1.110223e-16
                                                    
$double.xmin
[1] 2.225074e-308
 
$double.xmax
[1] 1.797693e+308

$double.base
[1] 2

$double.digits
[1] 53

$double.rounding
[1] 5

$double.guard
[1] 0

$double.ulp.digits
[1] -52

$double.neg.ulp.digits
[1] -53

$double.exponent
[1] 11

$double.min.exp
[1] -1022

$double.max.exp
[1] 1024

$integer.max
[1] 2147483647

$sizeof.long
[1] 4

$sizeof.longlong
[1] 8

$sizeof.longdouble
[1] 8

$sizeof.pointer
[1] 4

$sizeof.time_t
[1] 8

> capabilities()
       jpeg         png        tiff       tcltk         X11        aqua 
       TRUE        TRUE        TRUE        TRUE       FALSE       FALSE 
   http/ftp     sockets      libxml        fifo      cledit       iconv 
       TRUE        TRUE       FALSE        TRUE        TRUE        TRUE 
        NLS       Rprof     profmem       cairo         ICU long.double 
       TRUE        TRUE        TRUE        TRUE        TRUE       FALSE 
    libcurl 
       TRUE 
> 

[ben-schwen]

Are we sure the run in #1071737 includes #6137? From Dirks output it should be test_longdouble=FALSE and therefore #6137 should solve this problem. This would also explain why .Machine$longdouble.digits=NULL didn't trigger

[eddelbuettel]

The initial output (reported in Debian's bug tracker) used vanilla 1.15.0. I uploaded uesterday with the fix, builds worked (as expected) but I do not know what the cadence for the 'autopkgtest' machine is. I suspect it runs within a day.