You seem to be mixing up SDPParser.h and SDPContext.h -- but
MVFS file system are not posix compatible -- I ran into this when I
tried to diff two trees...
You can just use snapshot views?
(Or don't use clearcase ;-))
I suppose having a VPATH across two views will cause problems...
Supposedly this has been fixed:
(from gnu diff system.h:
/* Do struct stat *S, *T have the same file attributes?
POSIX says that two files are identical if st_ino and st_dev are
the same, but many filesystems incorrectly assign the same (device,
inode) pair to two distinct files, including:
- GNU/Linux NFS servers that export all local filesystems as a
single NFS filesystem, if a local device number (st_dev) exceeds
255, or if a local inode number (st_ino) exceeds 16777215.
- Network Appliance NFS servers in snapshot directories; see
Network Appliance bug #195.
- ClearCase MVFS; see bug id ATRia04618.
marty
Richard Sharman <[EMAIL PROTECTED]> writes on Sun, 13 Jan 2008 16:20:2
3 EST
>
> --w+3wPgj7G4
> Content-Type: text/plain; charset=us-ascii
> Content-Description: message body text
> Content-Transfer-Encoding: 7bit
>
> The Linux version of the MVFS file-system from Clearcase can present
> the same inode and dev numbers for a directory in different views.
> (We do not see the same problem on the sparc mvfs.)
>
> This can cause make to produce incorrect results when two or more
> clearcase view directories appear in the VPATH list.
>
> This little makefile demonstrates the problem:
>
> ------------------------------------------------------------------------
> VPATH := /view/sharman_Platform_Services/vobs
/view/sharman_MediaServer_1.0/vobs
>
> junk : junk.c NETSVC/NS_Inc/SDPContext.h
> wc -l $^
> ------------------------------------------------------------------------
>
> The file NETSVC/NS_Inc/SDPParser.h does not occur in the first vpath
> directory, but does occur in the second. (The directory NETSVC/NS_Inc
> exists in the first directory but is empty.)
>
> $ make -f make1
> make: *** No rule to make target `NETSVC/NS_Inc/SDPContext.h', needed
by `junk'. Stop.
> $
>
>
> I think that because both the directories have the same inode & dev
> combination, when make finds that the file does not occur in the first
> directory when it comes to the second one it thinks it has already
> processed it.
>
> The program filenumber does a stat and prints the inode number, the
> dev number and the ctime number. It demonstrates the fact that the
> two directories incorrectly have the same inode & dev numbers:
>
> $ V1=/view/sharman_Platform_Services/vobs/NETSVC/NS_Inc
> $ V2=/view/sharman_MediaServer_1.0/vobs/NETSVC/NS_Inc
> $ filenumber $V1 $V2
> /view/sharman_Platform_Services/vobs/NETSVC/NS_Inc device=64608
inode=3544 ctime=1032294808
> /view/sharman_MediaServer_1.0/vobs/NETSVC/NS_Inc device=64608
inode=3544 ctime=1177967845
> $ ls $V1/SDPContext.h $V2/SDPContext.h
> ls: /view/sharman_Platform_Services/vobs/NETSVC/NS_Inc/SDPContext.h: No
such file or directory
> /view/sharman_MediaServer_1.0/vobs/NETSVC/NS_Inc/SDPContext.h
> $
>
>
> I was able to overcome the problem by adding a check of the files'
> ctimes as well as their inode and dev numbers.
>
> $ ~/make-3.81/x86-linux/make -f make1
> wc -l junk.c /view/sharman_MediaServer_1.0/vobs/NETSVC/NS_Inc/SDPContext
.h
> 0 junk.c
> 31 /view/sharman_MediaServer_1.0/vobs/NETSVC/NS_Inc/SDPContext.h
> 31 total
> $
>
>
> Here is a patch against make 3.81
>
_______________________________________________
Bug-make mailing list
Bug-make@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-make
