R devel - Aug 2007 - fix for broken largefile seek() on 32-bit linux (PR#9883)

Full_Name: John Brzustowski
Version: R-devel-trunk, R-2.4.0
OS: linux
Submission from: (NULL) (206.248.132.197)


DESCRIPTION

seek() on files larger than 2 gigabytes fails for large values of
"where" on
i386 linux 2.6.13 (and presumably other 32-bit unix-like platforms).

e.g.:
> f<-file("3gigabytefile.dat", "rb")
> seek(f, 3e9, "start", "r")
[1] 0  ## correct
> seek(f, NA, "start", "r")
[1] 0  ## should be 3e+09

DIAGNOSIS

Typo: the compile-time tests for large file support use "HAVE_SEEKO"
instead of
"HAVE_FSEEKO", and so fail.  

The same typo appears in one of the extra/zlib files, so I'm fixing
it in the patch below, but I haven't tested whether that actually
produces a bug.

PATCH
Index: src/extra/zlib/gzio.c
==================================================================---
src/extra/zlib/gzio.c	(revision 42664)
+++ src/extra/zlib/gzio.c	(working copy)
@@ -25,7 +25,7 @@
 #include "zutil.h"
 
 /* R ADDITION */
-#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
+#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
 #define f_seek fseeko
 #define f_tell ftello
 #else
Index: src/include/Rconnections.h
==================================================================---
src/include/Rconnections.h	(revision 42664)
+++ src/include/Rconnections.h	(working copy)
@@ -63,7 +63,7 @@
 
 typedef struct fileconn {
     FILE *fp;
-#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
+#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
     off_t rpos, wpos;
 #else
 #ifdef Win32
Index: src/main/connections.c
==================================================================---
src/main/connections.c	(revision 42664)
+++ src/main/connections.c	(working copy)
@@ -446,7 +446,7 @@
 
 /* ------------------- file connections --------------------- */
 
-#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
+#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
 #define f_seek fseeko
 #define f_tell ftello
 #else
@@ -570,7 +570,7 @@
 {
     Rfileconn this = con->private;
     FILE *fp = this->fp;
-#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
+#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
     off_t pos;
 #else
 #ifdef Win32

I've applied the patch, but I no longer have a 32-bit Linux system. Could 
you please verify it has applied correctly (SVN rev 42672 on trunk, 42673 
on R-patched).  A quick check on 32-bit Solaris suggested it worked as 
expected.

On Mon, 27 Aug 2007, jbrzusto at fastmail.fm wrote:
> Full_Name: John Brzustowski
> Version: R-devel-trunk, R-2.4.0
> OS: linux
> Submission from: (NULL) (206.248.132.197)
>
>
> DESCRIPTION
>
> seek() on files larger than 2 gigabytes fails for large values of
"where" on
> i386 linux 2.6.13 (and presumably other 32-bit unix-like platforms).
>
> e.g.:
>
>> f<-file("3gigabytefile.dat", "rb")
>> seek(f, 3e9, "start", "r")
> [1] 0  ## correct
>> seek(f, NA, "start", "r")
> [1] 0  ## should be 3e+09
>
> DIAGNOSIS
>
> Typo: the compile-time tests for large file support use
"HAVE_SEEKO" instead of
> "HAVE_FSEEKO", and so fail.
>
> The same typo appears in one of the extra/zlib files, so I'm fixing
> it in the patch below, but I haven't tested whether that actually
> produces a bug.
>
> PATCH
> Index: src/extra/zlib/gzio.c
> ==================================================================> ---
src/extra/zlib/gzio.c	(revision 42664)
> +++ src/extra/zlib/gzio.c	(working copy)
> @@ -25,7 +25,7 @@
> #include "zutil.h"
>
> /* R ADDITION */
> -#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
> +#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
> #define f_seek fseeko
> #define f_tell ftello
> #else
> Index: src/include/Rconnections.h
> ==================================================================> ---
src/include/Rconnections.h	(revision 42664)
> +++ src/include/Rconnections.h	(working copy)
> @@ -63,7 +63,7 @@
>
> typedef struct fileconn {
>     FILE *fp;
> -#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
> +#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
>     off_t rpos, wpos;
> #else
> #ifdef Win32
> Index: src/main/connections.c
> ==================================================================> ---
src/main/connections.c	(revision 42664)
> +++ src/main/connections.c	(working copy)
> @@ -446,7 +446,7 @@
>
> /* ------------------- file connections --------------------- */
>
> -#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
> +#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
> #define f_seek fseeko
> #define f_tell ftello
> #else
> @@ -570,7 +570,7 @@
> {
>     Rfileconn this = con->private;
>     FILE *fp = this->fp;
> -#if defined(HAVE_OFF_T) && defined(HAVE_SEEKO)
> +#if defined(HAVE_OFF_T) && defined(HAVE_FSEEKO)
>     off_t pos;
> #else
> #ifdef Win32
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
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