[R-pkg-devel] Package valgrind problem I can't solve: Direction?

[Iñaki Úcar]

Hi,

2017-10-26 5:16 GMT+02:00 Peter Dunn <PDunn2 at usc.edu.au>:
> Hi all
>
> I am trying to compile (on my Mac) an R package (tweedie) which includes Fortran 77 code. I’m not much of a programmer, but can still manage to write and update F77 code. I’m new to valgrind.
>
> In checking my package (which passes the build-and-check with no errors), I use valgrind like this:
>
> R -d "valgrind  --tool=memcheck --leak-check=full --track-origins=yes" --vanilla  < tweedie-Ex.R
>
>
> …on my package example, which gives me a series of messages like this:
>
> ==53843== Conditional jump or move depends on uninitialised value(s)
> ==53843==    at 0x1118F61FB: smallp_ (in /Users/pdunn2/Library/R/3.4/library/tweedie/libs/tweedie.so)
> ==53843==    by 0xBFA4FB88FFFFFFEB: ???
> ==53843==    by 0x3FF91CB2DA3A50E7: ???
> ==53843==    by 0x104888757: ???
> ==53843==    by 0xFFFFFFEC03E375AF: ???
> ==53843==    by 0x3FE18C2EE1BE47CB: ???
> ==53843==    by 0x104885DEF: ???
> ==53843==    by 0x104885DF7: ???
> ==53843==    by 0x3FE518F314C67BE7: ???
> ==53843==    by 0x11364CEAF: ???
> ==53843==    by 0x104885DBF: ???
> ==53843==    by 0x100000007: ??? (in /Library/Frameworks/R.framework/Resources/bin/exec/R)
> ==53843==  Uninitialised value was created by a stack allocation
> ==53843==    at 0x1118F59A6: smallp_ (in /Users/pdunn2/Library/R/3.4/library/tweedie/libs/tweedie.so)

CRAN shows a complete backtrace:
https://www.stats.ox.ac.uk/pub/bdr/memtests/valgrind/tweedie/tweedie-Ex.Rout

According to it, the offending conditional jump seems to be this one:

         if (     ( its .LT. 3 )
     &         .OR.
     &            ( ( its .LT. maxit ) .AND.
     &              ( abs(relerr) .GT. aimrerr )
     &            )
     &      ) then

> I can see the piece of R code in the example from which this comes.  The subroutine  smallp  is mentioned in the second line of the output.
>
> I simply cannot find the issue.  (Spent days looking so far.)  There are only a few conditional jumps, and everything not passed to the subroutine  smallp  is initialised. I have checked the calls to  smallp  as well. I have compiled the fortran code separately (after wrapping into a complete program), with all kinds of flags:
>
> gfortran -O2 -fimplicit-none  -Wall  -Wline-truncation  -Wcharacter-truncation  -Wsurprising  -Waliasing  -Wunused-parameter  -fwhole-file  -fcheck=all  -std=f2008  -pedantic  -fbacktrace),

Did you try "-O0 -g" instead of "-O2"?

Regards,
Iñaki