bug-binutils
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug binutils/15545] New: BFD cache loses cloexec flag


From: tromey at redhat dot com
Subject: [Bug binutils/15545] New: BFD cache loses cloexec flag
Date: Tue, 28 May 2013 18:20:31 +0000

http://sourceware.org/bugzilla/show_bug.cgi?id=15545

            Bug ID: 15545
           Summary: BFD cache loses cloexec flag
           Product: binutils
           Version: unspecified
            Status: NEW
          Severity: minor
          Priority: P2
         Component: binutils
          Assignee: unassigned at sourceware dot org
          Reporter: tromey at redhat dot com

Created attachment 7047
  --> http://sourceware.org/bugzilla/attachment.cgi?id=7047&action=edit
test program

Consider the attached program.
(Compile it and run it on itself to see the error.)

It opens a BFD using an explicit fd.
Then it calls bfd_cache_close_all.

Despite the comment in opncls.c:

  /* If we opened the file by name, mark it cacheable; we can close it
     and reopen it later.  However, if a file descriptor was provided,
     then it may have been opened with special flags that make it
     unsafe to close and reopen the file.  */
  if (fd == -1)
    (void) bfd_set_cacheable (nbfd, TRUE);

... this causes the fd to be closed.  The BFD is not marked "cacheable"
but bfd_cache_close_all ignores this flag.

Then, the test program forces the BFD to be reopened.
Now, the "cacheable" flag is set.

All this is more subtle than it appears because I think GDB relies
on bfd_cache_close_all actually closing such fds.

One idea would be to simply remove the calls to bfd_cache_init from
opncls.c, but at the same time change gdb to add them.

-- 
You are receiving this mail because:
You are on the CC list for the bug.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]