Behavior of lseek() on a fd opened with 'RDONLY' flag, when seeking goes beyond file size.
Thanks David !
I have one more question, Does lseek() permits -ve value as offset, with SEEK_SET mode? When I execute a simple c program to check that it doesn't return any error.
errno = 0;
rd_rc = lseek(rd_fd, -10, SEEK_SET);
returns rd_rc = -10 and errno remains 0.
Quoted 1 line
From the linux code, I am expecting the return value of
-EINVAL,
when SEEK_SET is passed with negative offset (generic_file_llseek).
http://lxr.linux.no/linux/fs/read_write.c#L34
What is the expected behavior as per posix and linux?
Thanks,
Saravanan
----- Original Message ----
From: David Newall <[email protected]>
To: palani saravanan <[email protected]>
Cc: [email protected]
Sent: Tuesday, 17 June, 2008 8:00:09 PM
Subject: Re: Behavior of lseek() on a fd opened with 'RDONLY' flag, when seeking goes beyond file size.
Quoted 4 lines · palani saravanan
In linux, I see that it just goes beyond the file size and returns the resulting offset. For example, 'rc =
lseek(fd, 4L, SEEK_END);' on a file which has 5 byte contents, it returns rc as 9.
Sounds right. It's documented that way, too:
"The lseek() function allows the file offset to be set beyond the
end of the file (but this
does not change the size of the file)."
-- man 2 lseek
Quoted 2 lines
I expect that it would return size of the file, i.e.) 5.
Only if you pass (0,SEEK_END).
Quoted 2 lines
Does the file pointer internally really points to the new location.?
So the documentation promises.
Download prohibited? No problem. CHAT from any browser, without download. Go to http://in.messenger.yahoo.com/webmessengerpromo.php/
Quoted 5 lines · palani saravanan
Does
lseek()permits -ve value as offset, withSEEK_SETmode? ...errno = 0; rd_rc = lseek(rd_fd, -10, SEEK_SET);returns
rd_rc= -10 and errno remains 0.
No, lseek does not permit negative file offsets, and should return EINVAL in your example. You may have discovered a bug. For reference, on my system lseek returns -1 and sets errno to EINVAL, as it should.
If you're running a recent kernel it might be worth looking closer into this; for older kernels it's probably not worth the effort.