JehanneOS/newlib
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

* DevNotes: Add entry cgf-000011.

Browse Source 
* fhandler.h (fhandler_base::refcnt): Delete.
(fhandler_base::inc_refcnt): New function.
(fhandler_base::dec_refcnt): New function.
* cygheap.h (cygheap_fdnew::~cygheap_fdnew): Accommodate split of refcnt to
inc_refcnt/dec_refcnt.
(cygheap_fdget::cygheap_fdget): Ditto.
(cygheap_fdget::~cygheap_fdget::cygheap_fdget): Ditto.
* dtable.cc (dtable::release): Ditto.
(cygwin_attach_handle_to_fd): Ditto.
(dtable::init_std_file_from_handle): Ditto.
(dtable::dup3): On success, return with fdtab locked.
* dtable.h (dtable): Add dup_finish as a friend.
* syscalls.cc (dup_finish): Define new function.  Increment refcnt while fdtab
is locked.
(dup2): Use common dup_finish() to perform dup operation.
(dup3): Ditto.
This commit is contained in:
Christopher Faylor
2012-06-03 18:02:45 +00:00
parent ff80d22a7c
commit 3143cb7c00
7 changed files with 74 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
winsup/cygwin/fhandler.h
View File

@ -182,15 +182,8 @@ class fhandler_base
HANDLE read_state;
public:
long refcnt(long i = 0)
{
debug_only_printf ("%p, %s, i %d, refcnt %ld", this, get_name (), i, _refcnt + i);
/* This MUST be an interlocked operation. If multiple threads access the
same descriptor in quick succession, a context switch can interrupt
the += operation and we wrongly end up with a refcnt of 0 in the
cygheap_fdget destructor. */
return i ? InterlockedAdd (&_refcnt, i) : _refcnt;
}
long inc_refcnt () {return InterlockedIncrement (&_refcnt);}
long dec_refcnt () {return InterlockedDecrement (&_refcnt);}
class fhandler_base *archetype;
int usecount;