Austin Group Defect Tracker

Aardvark Mark IV

Viewing Issue Simple Details Jump to Notes ] Issue History ] Print ]
ID Category Severity Type Date Submitted Last Update
0000480 [1003.1(2008)/Issue 7] System Interfaces Objection Error 2011-07-28 15:08 2019-06-10 08:55
Reporter eblake View Status public  
Assigned To ajosey
Priority normal Resolution Accepted As Marked  
Status Closed  
Name Eric Blake
Organization Red Hat
User Reference ebb.handles
Section 2.5.1 Interaction of File Descriptors and Standard I/O Streams
Page Number 491
Line Number 16844
Interp Status Approved
Final Accepted Text See Note: 0000894
Summary 0000480: freopen, fclose effect on multiple file handles
Description The standard states that freopen always closes a handle and creates a
stream that is open to a new file description when describing the
interaction of freopen with handles, but later states that freopen
with a NULL filename argument may (but not must) reuse the existing
stream and underlying file description [XSH freopen lines 30885-89].

Here, I argue that the freopen description is intended, and that the
multiple-handle text should be fixed to be made consistent.

Also, the following example of two handles should be well-defined, but
is currently handled differently in Solaris than it is in glibc 2.14;
I argue that the glibc behavior is a bug in trying to implement 0000087
(that is, in trying to fix fclose() vs. ungetc() semantics in glibc bug, [^] glibc inadvertently
also broke fclose() to blindly reposition the file description even if
the stream was not otherwise used):

  const char buf[] = "hello world";
  int fd;
  int fd2;
  FILE *f;

  /* Prepare a seekable file. */
  fd = open (NAME, O_RDWR | O_CREAT | O_TRUNC, 0600);
  assert (0 <= fd);
  assert (write (fd, buf, sizeof buf) == sizeof buf);
  assert (lseek (fd, 1, SEEK_SET) == 1);

  /* Create an output stream visiting the file; when it is closed, all
     other file descriptors visiting the file must see the new file
     position. */
  fd2 = dup (fd);
  assert (0 <= fd2);
  f = fdopen (fd2, "w"); // POSIX requires that f be at position 1
  assert (f);
  assert(lseek(fd, 4, SEEK_SET) == 4); // fd is now the active handle
  assert (fclose (f) == 0); // The f/fd2 handle was never repositioned or made active
  assert (lseek (fd, 0, SEEK_CUR) == 4); // so fd should still be at 4

Solaris passed this test, but glibc 2.14 ended up repositioning the
file description to offset 1, which was then observable in fd and
causes the assertion to fail.

I think the intent of the standard is clear, but recent wording changes
to fclose() mean that we can make this more explicit.
Desired Action At line 16844 [XSH 2.5.1], change:

A handle which is a stream is considered to be closed when either an
fclose( ) or freopen( ) is executed on it (the result of freopen( ) is
a new stream, which cannot be a handle on the same open file description
as its previous value),


A handle which is a stream is considered to be closed when either an
fclose( ), or freopen( ) with non-null filename, is executed on it (for
freopen( ) with a null filename, it is implementation-defined whether
a new handle is created or the existing one reused),

At line 16880, add a bullet for the second handle:

• If fclose( ) is the first (and only) action on the second handle, then
the offset of the file description shall not be changed from what it was
via the first handle.
Tags tc2-2008
Attached Files

- Relationships
related to 0000087Closedajosey fflush and ungetc 
related to 0000662Under Reviewajosey Clarify or add file descriptor preservation and atomicity requirements for freopen 

-  Notes
jim_pugsley (manager)
2011-07-28 16:27

Interpretation response
The standard is unclear on this issue, and no conformance distinction can be made between alternative implementations based on this. This is being referred to the sponsor.

Standard is contradictory on whether freopen() can reuse a file description when passed a null filename.

Notes to the Editor (not part of this interpretation):
Make the changes in the suggested action except do not make the changes suggested at line 16880.
ajosey (manager)
2011-08-10 15:19

Interpretation proposed 10 August 2011 for final 30 day review
ajosey (manager)
2011-09-12 15:21

Interpretation approved 12 Sept 2011

- Issue History
Date Modified Username Field Change
2011-07-28 15:08 eblake New Issue
2011-07-28 15:08 eblake Status New => Under Review
2011-07-28 15:08 eblake Assigned To => ajosey
2011-07-28 15:08 eblake Name => Eric Blake
2011-07-28 15:08 eblake Organization => Red Hat
2011-07-28 15:08 eblake User Reference => ebb.handles
2011-07-28 15:08 eblake Section => 2.5.1 Interaction of File Descriptors and Standard I/O Streams
2011-07-28 15:08 eblake Page Number => 491
2011-07-28 15:08 eblake Line Number => 16844
2011-07-28 15:08 eblake Interp Status => ---
2011-07-28 16:22 jim_pugsley Interp Status --- => Pending
2011-07-28 16:22 jim_pugsley Status Under Review => Interpretation Required
2011-07-28 16:22 jim_pugsley Resolution Open => Accepted As Marked
2011-07-28 16:27 jim_pugsley Note Added: 0000894
2011-07-28 16:27 jim_pugsley Tag Attached: tc2-2008
2011-07-28 16:29 jim_pugsley Final Accepted Text => See Note: 0000894
2011-07-28 16:31 eblake Relationship added related to 0000087
2011-08-10 15:19 ajosey Interp Status Pending => Proposed
2011-08-10 15:19 ajosey Note Added: 0000927
2011-09-12 15:21 ajosey Interp Status Proposed => Approved
2011-09-12 15:21 ajosey Note Added: 0000959
2013-02-21 18:04 eblake Relationship added related to 0000662
2019-06-10 08:55 agadmin Status Interpretation Required => Closed

Mantis 1.1.6[^]
Copyright © 2000 - 2008 Mantis Group
Powered by Mantis Bugtracker