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

* cygtls.cc (_cygtls::init_exception_handler): Force installation of our 

exception handler to always be at the beginning.
This commit is contained in:
Christopher Faylor 2010-02-24 00:03:42 +00:00
parent 84f66e46b9
commit fd4a56afe8
2 changed files with 29 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
winsup/cygwin/ChangeLog
View File

@ -1,3 +1,8 @@
2010-02-23 Christopher Faylor <me+cygwin@cgf.cx>
* cygtls.cc (_cygtls::init_exception_handler): Force installation of
our exception handler to always be at the beginning.
2010-02-23 Christopher Faylor <me+cygwin@cgf.cx> 2010-02-23 Christopher Faylor <me+cygwin@cgf.cx>
* thread.cc (pthread_mutex::unlock): Don't attempt to unlock if there * thread.cc (pthread_mutex::unlock): Don't attempt to unlock if there

31
winsup/cygwin/cygtls.cc
View File

@ -240,14 +240,30 @@ _cygtls::init_exception_handler (exception_handler *eh)
a problem if some previous exception handler tries to do things that are a problem if some previous exception handler tries to do things that are
better left to Cygwin. I await the cygwin mailing list notification of better left to Cygwin. I await the cygwin mailing list notification of
this event with bated breath. this event with bated breath.
(cgf 2009-07-17)
(cgf 2009-07-17) */ A change in plans: In the not-so-distant past of 2010-02-23 it was
for (exception_list *e = _except_list; discovered that something was moving in ahead of cygwin's exception
e != NULL && e != (exception_list *) -1; handler so just detecting that the exception handler was loaded wasn't
e = e->prev) good enough. I sort of anticipated this. So, the next step is to remove
if (e == &el) the old exception handler from the list and add it to the beginning.
return;
el.handler = eh; The next step will probably be to call this function at various points
in cygwin (like from _cygtls::setup_fault maybe) to absoltely ensure that
we have control. For now, however, this seems good enough.
(cgf 2010-02-23)
*/
exception_list *e = _except_list;
if (e == &el)
return;
while (e && e != (exception_list *) -1)
if (e->prev != &el)
e = e->prev;
else
{
e->prev = el.prev;
break;
}
/* Apparently Windows stores some information about an exception and tries /* Apparently Windows stores some information about an exception and tries
to figure out if the SEH which returned 0 last time actually solved the to figure out if the SEH which returned 0 last time actually solved the
problem, or if the problem still persists (e.g. same exception at same problem, or if the problem still persists (e.g. same exception at same
@ -259,6 +275,7 @@ _cygtls::init_exception_handler (exception_handler *eh)
Windows 2008, which irremediably gets into an endless loop, taking 100% Windows 2008, which irremediably gets into an endless loop, taking 100%
CPU. That's why we reverted to a normal SEH chain and changed the way CPU. That's why we reverted to a normal SEH chain and changed the way
the exception handler returns to the application. */ the exception handler returns to the application. */
el.handler = eh;
el.prev = _except_list; el.prev = _except_list;
_except_list = &el; _except_list = &el;
} }