On Jul  8 12:44, Christopher Faylor wrote:
> On Tue, Jul 08, 2014 at 11:36:07AM -0500, Yaakov Selkowitz wrote:
> >On 2014-07-08 10:02, Christopher Faylor wrote:
> >> On Tue, Jul 08, 2014 at 09:55:55AM -0400, Ken Brown wrote:
> >>> Grasping at straws, as usual, I wonder if these mysterious crashes could
> >>> be related to a bug that Corinna just fixed:
> >>>
> >>>    https://cygwin.com/ml/cygwin-cvs/2014-q3/msg00004.html
> >>>
> >>> Corinna, is this plausible?  If so, maybe Gustav should try the next
> >>> snapshot of the Cygwin DLL.  (The current one doesn't seem to have this
> >>> fix in it.)
> >>
> >> Actually, I was scratching my head over that change and wondering
> >> what it was supposed to solve.
> >
> >We believe that this was the cause of e.g. mandb aborting on x86_64 
> >without manually enlarging the stack commit size:
> 
> "We"?

Yaakov and me in a discussion with integrated testing on IRC.  What are
you trying to imply?

> Isn't that testable by setting the stack size down and putting a
> printf in cygwin somewhere?
> 
> In any event, the comment in my code which is supposed to explain why
> the tag is commented out is lacking details.  If this patch is a joint
> effort then please add more details to the change that you made to my
> code.

It's not your code.  The function exception::myfault_handle has been
written by me to workaround the lack of structured exception handling in
GCC on x86_64.

I don't understand what's unclear in this comment.  exception::myfault_handle
is a vectored exception handler (see the comment preceeding the method)
so it short-circuits structured exception handling.  Thus it short-circuits
PAGE_GUARD handling, aka stack commits.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat