[Rd] stack imbalance in max.col for non-real matrices

[Martin Morgan]

On 04/19/2013 01:16 AM, Prof Brian Ripley wrote:
> On 17/04/2013 21:25, Michael Lawrence wrote:
>> It's tough to reliably reproduce, but I often get stack imbalance warnings
>> when calling max.col() on non-real/double matrix. The code is conditionally
>> PROTECTing but not incrementing its nprot counter for the eventual
>> UNPROTECT. Pretty sure this would fix (but I haven't tested it):
>
> I can't reproduce it either, but that seems the right fix so incorporated now,
> thanks.

Why isn't this reproducible? I thought the protection stack would be deterministic.

 > max.col(matrix())
Warning: stack imbalance in 'max.col', 10 then 11
Warning: stack imbalance in '.Internal', 9 then 10
Warning: stack imbalance in '{', 6 then 7
[1] NA

 > sessionInfo()
R Under development (unstable) (2013-04-18 r62615)
Platform: x86_64-unknown-linux-gnu (64-bit)

from config.log

   $ /home/mtmorgan/src/R-devel/configure --enable-memory-profiling 
--disable-byte-compiled-packages --enable-R-shlib CC=clang CFLAGS=-g -O0 
CXX=clang++ CXXFLAGS=-g -O0 --no-create --no-recursion

with

$ clang --version
Ubuntu clang version 3.0-6ubuntu3 (tags/RELEASE_30/final) (based on LLVM 3.0)
Target: x86_64-pc-linux-gnu
Thread model: posix


>
>>
>> Index: array.c
>> ===================================================================
>> --- array.c    (revision 61606)
>> +++ array.c    (working copy)
>> @@ -1614,7 +1614,10 @@
>>       SEXP m = CAR(args);
>>       int method = asInteger(CADR(args));
>>       int nr = nrows(m), nc = ncols(m), nprot = 1;
>> -    if (TYPEOF(m) != REALSXP) PROTECT(m = coerceVector(m, REALSXP));
>> +    if (TYPEOF(m) != REALSXP) {
>> +      PROTECT(m = coerceVector(m, REALSXP));
>> +      nprot++;
>> +    }
>>       SEXP ans = allocVector(INTSXP, nr);
>>       PROTECT(ans);
>>       R_max_col(REAL(m), &nr, &nc, INTEGER(ans), &method);
>>
>> Thanks,
>> Michael
>>
>>     [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>
>


-- 
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109

Location: Arnold Building M1 B861
Phone: (206) 667-2793