rsync - May 2023 - xfer_sum_len type bug

Hi folks,

I have several build environments in which I must build rsync, and in
some, but not all of them, the build fails when built with -Wall
-Werror due to this warning:

In file included from /usr/include/string.h:495,
                 from
/home/demartin/BuildClients/cobrasync-8.80-alsi11-lib32/cobrasync/rsync-3.2.7/rsync.h:339,
                 from
/home/demartin/BuildClients/cobrasync-8.80-alsi11-lib32/cobrasync/rsync-3.2.7/match.c:22:
In function 'memset',
    inlined from 'match_sums' at
/home/demartin/BuildClients/cobrasync-8.80-alsi11-lib32/cobrasync/rsync-3.2.7/match.c:431:3:
/usr/include/i386-linux-gnu/bits/string_fortified.h:71:10: error:
'__builtin___memset_chk' specified size between 2147483648 and
4294967295 exceeds maximum object size 2147483647 [-Werror=stringop-overflow=]
   71 |   return __builtin___memset_chk (__dest, __ch, __len, __bos0 (__dest));
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors

This appears to be because of a type mismatch between xfer_sum_len
(declared as [signed] int) and the third arugment to memset, whose
function prototype is (from the man page):

       void *memset(void *s, int c, size_t n);
  
size_t != int.

I did check, and it appears to me that the other places xfer_sum_len
is used as a function argument expect a size_t.  Assuming I haven't
missed any important details, the fix is simple enough (and does
indeed compile for me where the original does not):

---------8K----------
diff -Naur rsync-3.2.7.patch_orig/checksum.c rsync-3.2.7.patched/checksum.c
--- rsync-3.2.7.patch_orig/checksum.c	2023-05-16 14:17:31.078521657 -0400
+++ rsync-3.2.7.patched/checksum.c	2023-05-16 15:36:44.253892249 -0400
@@ -98,7 +98,7 @@
     { CSUM_MD5, NNI_BUILTIN, "md5", NULL };
 
 struct name_num_item *xfer_sum_nni; /* used for the transfer checksum2
computations */
-int xfer_sum_len;
+size_t xfer_sum_len;
 struct name_num_item *file_sum_nni; /* used for the pre-transfer --checksum
computations */
 int file_sum_len, file_sum_extra_cnt;
 
diff -Naur rsync-3.2.7.patch_orig/match.c rsync-3.2.7.patched/match.c
--- rsync-3.2.7.patch_orig/match.c	2023-05-16 14:17:31.082521665 -0400
+++ rsync-3.2.7.patched/match.c	2023-05-16 15:36:46.781895867 -0400
@@ -32,7 +32,7 @@
 extern int append_mode;
 
 extern struct name_num_item *xfer_sum_nni;
-extern int xfer_sum_len;
+extern size_t xfer_sum_len;
 
 int updating_basis_file;
 char sender_file_sum[MAX_DIGEST_LEN];
diff -Naur rsync-3.2.7.patch_orig/receiver.c rsync-3.2.7.patched/receiver.c
--- rsync-3.2.7.patch_orig/receiver.c	2023-05-16 14:17:31.082521665 -0400
+++ rsync-3.2.7.patched/receiver.c	2023-05-16 15:37:10.401929552 -0400
@@ -75,7 +75,7 @@
 extern OFF_T preallocated_len;
 
 extern struct name_num_item *xfer_sum_nni;
-extern int xfer_sum_len;
+extern size_t xfer_sum_len;
 
 static struct bitbag *delayed_bits = NULL;
 static int phase = 0, redoing = 0;
---------8K----------


-- 
Derek Martin
Principal System Software Engineer
Akamai Technologies
demartin at akamai.com

On Tue, May 16, 2023 at 2:03?PM Derek Martin wrote:
> This appears to be because of a type mismatch between xfer_sum_len
> (declared as [signed] int) and the third arugment to memset, whose
> function prototype is (from the man page):
>
>        void *memset(void *s, int c, size_t n);
>
If that is the case, why does it not complain about file_sum_len, which is
also an int that is passed to memset()? The only difference I see is that
xfer_sum_len's memset() is in match.c while file_sum_len's memset() is
in
checksum.c, and the only difference I can see with that is that the int is
local to checksum.c and extern in match.c. Very strange. I'm more inclined
to change them into unsigned short values, as they are quite small integers.

..wayne..
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.samba.org/pipermail/rsync/attachments/20230526/47967305/attachment.htm>