Everyone knows the truncate(file, size) function, which changes the file size to a given size, by truncating the tail of the file. But how to do the same, only with the truncation of not the tail of the file and his head?
Asked
Active
Viewed 2,130 times
4
Vyacheslav
- 171
- 1
- 8
-
When the file is a logfile and the program writing to it keeps the filedescriptor open, no more logging will appear after moving the file. See https://unix.stackexchange.com/a/122942/57293 – Walter A Feb 11 '19 at 09:04
-
open the file, then `sendfile(fd, fd, &offset, size)`, I test on ubuntu 16, works ok, but the older kernel doesn't support – yuanjianpeng Jul 03 '19 at 06:47
3 Answers
4
Generally, you have to rewrite the entire file. The simplest way is to skip the first few bytes, copy everything else to a temporary file, and rename the temporary file on top of the old one when done. A more elaborate way is to rewrite it in place, analogous to how memmove works, with read/seek/write/seek or pread/pwrite, and then truncate the last bit when done.
If you are on a recent version of Linux (>= 3.15), and you have a supported filesystem (currently ext4 or xfs), and the amount you wish to remove happens to be a multiple of the filesystem block size, you could use the non-portable fallocate(2) with the FALLOC_FL_COLLAPSE_RANGE flag. Note that this feature is not supported by the portable posix_fallocate.
Nate Eldredge
- 48,811
- 6
- 54
- 82
3
You could just use tail --lines=<linecount> to always cap the log file to the last linecount lines. This works if you're not trying to truncate to a specific / fixed file size.
cyclone72
- 31
- 1
2
The only way is to read the whole file and overwrite it from a needed position.
montonero
- 1,363
- 10
- 16