Bill,
Thanks. Whether or not it is the cause here it should clearly be
PROTECT-ed (I suspect the original version didn't need to be, but leaving
PROTECTs off is prejudicial to future maintenance), so I've incorporated
this in R-devel/R-patched.
Brian
On Thu, 7 Aug 2008, Bill Dunlap wrote:
On Thu, 7 Aug 2008 [EMAIL PROTECTED] wrote:
Full_Name: Juan Gea
Version: R version 2.6.2
OS: Fedora Core 6
Submission from: (NULL) (79.153.48.49)
Abort:
objeS <- matrix("AAA",1000000)
class(objeS)
outTxt <- textConnection("vaClob", open = "w", local = FALSE)
dput(objeS,outTxt)
close(outTxt)
R version 2.6.2 (2008-02-08)
...
objeS <- matrix("AAA",1000000)
class(objeS)
[1] "matrix"
outTxt <- textConnection("vaClob", open = "w", local = FALSE)
dput(objeS,outTxt)
*** caught segfault ***
address 0xb7803934, cause 'memory not mapped'
Traceback:
1: dput(objeS, outTxt)
Does the following solve the problem? valgrind on R 2.8.0 finds
deparse.c:do_dput() reading from freed memory, tval, the output
of deparse1(), right after a call to Rconn_printf() invokes the
garbage collector and frees tval.
It is taking a long time to run the example under valgrind
after the fix, but I think it has gotten past the original error.
Index: src/main/deparse.c
===================================================================
--- src/main/deparse.c (revision 46243)
+++ src/main/deparse.c (working copy)
@@ -284,7 +284,7 @@
if(!isNull(CADDR(args)))
opts = asInteger(CADDR(args));
- tval = deparse1(tval, 0, opts);
+ PROTECT(tval = deparse1(tval, 0, opts));
if (TYPEOF(CAR(args)) == CLOSXP) {
SET_CLOENV(CAR(args), saveenv);
UNPROTECT(1);
@@ -313,6 +313,7 @@
res < strlen(CHAR(STRING_ELT(tval, i))) + 1)
warning(_("wrote too few characters"));
}
+ UNPROTECT(1) ; /* tval */
if (!wasopen) con->close(con);
return (CAR(args));
}
The valgrind session was:
> objeS <- matrix("AAA",1000000)
> class(objeS)
[1] "matrix"
> outTxt <- textConnection("vaClob", open = "w", local = FALSE)
> dput(objeS,outTxt)
==32381== Invalid read of size 4
==32381== at 0x813917E: do_dput (deparse.c:312)
==32381== by 0x8065454: do_internal (names.c:1138)
==32381== by 0x8160557: Rf_eval (eval.c:461)
==32381== by 0x8160557: Rf_eval (eval.c:461)
==32381== by 0x8162267: do_begin (eval.c:1174)
==32381== by 0x8160557: Rf_eval (eval.c:461)
==32381== by 0x81636D0: Rf_applyClosure (eval.c:667)
==32381== by 0x816048F: Rf_eval (eval.c:505)
==32381== by 0x8058033: Rf_ReplIteration (main.c:257)
==32381== by 0x805825E: R_ReplConsole (main.c:306)
==32381== by 0x80584F4: run_Rmainloop (main.c:966)
==32381== by 0x805674D: main (Rmain.c:33)
==32381== Address 0x59B9B58 is 2,864 bytes inside a block of size 444,472
free'd
==32381== at 0x40052A3: free (vg_replace_malloc.c:233)
==32381== by 0x805ACFD: R_gc_internal (memory.c:767)
==32381== by 0x805BA4A: Rf_allocVector (memory.c:1971)
==32381== by 0x810C45C: Rf_lengthgets (builtin.c:746)
==32381== by 0x8123B0D: text_vfprintf (connections.c:2303)
==32381== by 0x8127147: Rconn_printf (connections.c:2868)
==32381== by 0x813917D: do_dput (deparse.c:311)
==32381== by 0x8065454: do_internal (names.c:1138)
==32381== by 0x8160557: Rf_eval (eval.c:461)
==32381== by 0x8160557: Rf_eval (eval.c:461)
==32381== by 0x8162267: do_begin (eval.c:1174)
==32381== by 0x8160557: Rf_eval (eval.c:461)
----------------------------------------------------------------------------
Bill Dunlap
Insightful Corporation
bill at insightful dot com
360-428-8146
"All statements in this message represent the opinions of the author and do
not necessarily reflect Insightful Corporation policy or position."
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Brian D. Ripley, [EMAIL PROTECTED]
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
