[debbugs-tracker] bug#22769: closed (In-place sorting of a file)

[GNU bug Tracking System]

Your message dated Mon, 22 Feb 2016 23:53:25 -0800
with message-id <address@hidden>
and subject line Re: bug#22769: In-place sorting of a file
has caused the debbugs.gnu.org bug report #22769,
regarding In-place sorting of a file
to be marked as done.

(If you believe you have received this mail in error, please contact
address@hidden)


-- 
22769: http://debbugs.gnu.org/cgi/bugreport.cgi?bug=22769
GNU Bug Tracking System
Contact address@hidden with problems

--- Begin Message --- Subject: In-place sorting of a file Date: Mon, 22 Feb 2016 04:57:54 -0500

Hi,

I found someone talking about sorting a file in-place:

http://stackoverflow.com/questions/9117274/do-i-need-to-generate-a-second-file-to-sort-a-file

However, sort file.txt -o file.txt is NOT safe. A strace sequence (sorting 1 2 3 4 5):

read(3, "1\n2\n3\n4\n5\n\n", 4096)      = 11

read(3, "", 4096)                       = 0

_llseek(3, 0, [11], SEEK_CUR)           = 0

close(3)                                = 0

munmap(0xb7743000, 4096)                = 0

ftruncate64(1, 0)                       = 0             <----------------------  (*)

fstat64(1, {st_mode=S_IFREG|0664, st_size=0, ...}) = 0

mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7743000

write(1, "\n1\n2\n3\n4\n5\n", 11)       = 11

close(1)                                = 0

If at this time, the system crashes (or later at write), the data is gone forever (or have a suffering recovery). Should we warn users of this, or add special treatment with this case (write a temp file and rename it)?

Thank your for your attention.

Regards,

Yanyan Jiang 蒋炎岩

Institute of Computer Software,

Dept. of Computer Science, Nanjing University

--- End Message ---

--- Begin Message --- Subject: Re: bug#22769: In-place sorting of a file Date: Mon, 22 Feb 2016 23:53:25 -0800 User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.5.1

Yanyan Jiang wrote:

If at this time, the system crashes (or later at write), the data is gone 
forever (or have a suffering recovery). Should we warn users of this, or add 
special treatment with this case (write a temp file and rename it)?

Thanks for reporting the issue. Writing to a temp file and renaming would have problems, as it might change the file's ownership or permissions or etc., and also won't behave as the user expects if the file has multiple hard links. We should at least warn users about the problem in the documentation, though, so I installed the attached patch on Savannah at:

http://git.savannah.gnu.org/cgit/coreutils.git/commit/?id=4c1995f1b673b0e1eae73a9db4e28016b86dee9e

0001-doc-document-sort-o-F-F-danger.txt
Description: Text document

--- End Message ---