Re: New tar available for testing (was Re: here is a patch for gnu tar incremental backup...)

Re: New tar available for testing (was Re: here is a patch for gnu tar incremental backup...)

[Jeremy Green]

On Sat, Nov 8, 2003 cgf wrote:

> Thanks for the patch but I don't think it is quite right and I don't
> think it catches everything.
>
> I'm uploading a new version of tar for testing now.  I'd appreciate
> feedback (to the cygwin list) on whether it solves the reported problem.

The binary you've uploaded, and the binary I built from the new source
exhibit the same problem that tar-1.13.25-3 showed:

$ touch f1
$ mkdir foo
$ touch foo/bar
$ ls -lR files/
files/:
total 0
-rw-r--r--    1 jeremy    None            0 Nov 10 12:26 f1
drwxr-xr-x+   2 jeremy    None            0 Nov 10 12:26 foo

files/foo:
total 0
-rw-r--r--    1 jeremy    None            0 Nov 10 12:26 bar
$ /usr/src/tar-1.13.25-4/src/tar --verbose --listed-incremental=list \
-cf archive.tar files
/usr/src/tar-1.13.25-4/src/tar: files/foo: Directory is new
files/
files/foo/
files/f1
files/foo/bar
$ /usr/src/tar-1.13.25-4/src/tar --verbose
--listed-incremental=list -cf archive.tar files
/usr/src/tar-1.13.25-4/src/tar: files/foo: Directory has been renamed
files/
files/foo/
files/foo/bar

I.e. files/foo/bar is backed-up even though it hasn't changed. With the
application of the following patch to the tar-1.13.25-4 source package:

--- src/incremen.c.orig 2003-11-10 12:27:36.094206400 +0000
+++ src/incremen.c      2003-11-10 12:33:06.629492800 +0000
@@ -437,7 +437,7 @@ read_directory_file (void)

          errno = 0;
 #ifdef __CYGWIN_USE_BIG_TYPES__
-         ino = strtoul (strp, &ebuf, 10);
+         ino = strtoull (strp, &ebuf, 10);
 #else
          ino = strtoul (strp, &ebuf, 10);
 #endif

The test works as expected:

$ rm list
$ /usr/src/tar-1.13.25-4/src/tar --verbose \
--listed-incremental=list -cf archive.tar files
/usr/src/tar-1.13.25-4/src/tar: files/foo: Directory is new
files/
files/foo/
files/f1
files/foo/bar
$ cat list
1068469530
+3160087061 8677873531989524896 files/foo
$ /usr/src/tar-1.13.25-4/src/tar --verbose \
--listed-incremental=list -cf archive.tar files
files/
files/foo/

However, if I use the tar-1.13.25-1 binary using this list file, I get...

$ cygcheck -c tar
Cygwin Package Information
Package              Version        Status
tar                  1.13.25-1      OK
$ tar --verbose --listed-incremental=list -cf archive.tar files
tar: list:2: Device number out of range
tar: list:2: Inode number out of range
tar: files/foo: Directory has been renamed
files/
files/foo/
files/foo/bar
tar: Error exit delayed from previous errors
$ tar --verbose --listed-incremental=list -cf archive.tar files
files/
files/foo/
ichthus$ cat list
1068469660
23317 2020475296 files/foo

i.e. different inode and device numbers are stored in the list file by
tar-1.13.25-1 and the patched version of tar-1.13.25-4. This doesn't
really matter as far as I'm concerned.

Jeremy

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

Re: New tar available for testing (was Re: here is a patch for gnu tar incremental backup...)

[Christopher Faylor]

On Mon, Nov 10, 2003 at 01:25:02PM +0000, Jeremy Green wrote:
>I.e. files/foo/bar is backed-up even though it hasn't changed. With the
>application of the following patch to the tar-1.13.25-4 source package:
>
>--- src/incremen.c.orig 2003-11-10 12:27:36.094206400 +0000
>+++ src/incremen.c      2003-11-10 12:33:06.629492800 +0000
>@@ -437,7 +437,7 @@ read_directory_file (void)
>
>          errno = 0;
> #ifdef __CYGWIN_USE_BIG_TYPES__
>-         ino = strtoul (strp, &ebuf, 10);
>+         ino = strtoull (strp, &ebuf, 10);
> #else
>          ino = strtoul (strp, &ebuf, 10);
> #endif

Odd.  I already have that change in my sandbox already but I must have
made the change after uploading the file.  I'll refresh the source and
binaries shortly.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

Available for testing: tar-1.13.25-5

[Christopher Faylor]

On Mon, Nov 10, 2003 at 09:25:35AM -0500, Christopher Faylor wrote:
>On Mon, Nov 10, 2003 at 01:25:02PM +0000, Jeremy Green wrote:
>>I.e. files/foo/bar is backed-up even though it hasn't changed. With the
>>application of the following patch to the tar-1.13.25-4 source package:
>>
>>--- src/incremen.c.orig 2003-11-10 12:27:36.094206400 +0000
>>+++ src/incremen.c      2003-11-10 12:33:06.629492800 +0000
>>@@ -437,7 +437,7 @@ read_directory_file (void)
>>
>>          errno = 0;
>> #ifdef __CYGWIN_USE_BIG_TYPES__
>>-         ino = strtoul (strp, &ebuf, 10);
>>+         ino = strtoull (strp, &ebuf, 10);
>> #else
>>          ino = strtoul (strp, &ebuf, 10);
>> #endif
>
>Odd.  I already have that change in my sandbox already but I must have
>made the change after uploading the file.  I'll refresh the source and
>binaries shortly.

The new version of tar is uploaded.  It should be on mirrors shortly.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/

New tar available for testing (was Re: here is a patch for gnu tar incremental backup...)

[Christopher Faylor]

On Mon, Oct 27, 2003 at 06:45:32AM -0500, George Carrette wrote:
>Here is a fix for gnu tar incremental backup feature
>for the most recent cygwin that uses a 64-bit value for 
>the type ion_t
>
>This was done against the sources in release/tar/tar-1.13.25-3-src.tar.bz2

Thanks for the patch but I don't think it is quite right and I don't think
it catches everything.

I'm uploading a new version of tar for testing now.  I'd appreciate feedback
(to the cygwin list) on whether it solves the reported problem.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/