This is the mail archive of the mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Signal handling tune up.

On Tue, Aug 19, 2003 at 08:46:36AM -0400, Pierre A. Humblet wrote:
>At 12:13 AM 8/19/2003 -0400, Christopher Faylor wrote:
>>I may not be able to get to your patch soon since I have actual
>>honest-to-gosh Red Hat Cygwin duties to attend to for the first time
>>in quite some time.  I volunteered to help out on a cygwin project even
>>though I've moved from the group that is responsible for cygwin.  I
>>neglected those duties tonight, in fact, thanks to the excitement of
>>someone actually submitting substantial signal handling patches.  :-)
>No problem, sorry to keep you awake! Some thoughts came to my mind 
>during the night, don't know if they are still relevant after your
>>Are you saying that that someone reported execing a process, hitting
>>ctrl-c, and having another process simultaneously sending CTRL-C's to
>>the exec'ed stub?  
>The thisproc argument in sig_handle was only used to handle that exec'ed
>stub. Couldn't that condition be detected reliably in handle_exceptions(),

Moot point but it used to be handled that way.  I don't remember why I
changed it.

>>I only applied the reversal of the movl with the call above since I'm
>>not convinced that moving the set_process_mask into interrupt_setup
>>doesn't introduce a race. 
>Don't we have the same problem today?
>Handler is running with current mask M1, old mask M0
>New signal occurs, sigthread prepares sigsave with new mask M2, old mask M1
> but is preempted before setting sigsave.sig
>Handler terminates, restores M0
>sigthread resumes and starts new handler. It runs with M2 and restores
>M1 (instead of M0) at end.

Is this any different from UNIX?  Some races in signal handling are
inavoidable, IMO.  I guess you could make things slightly more
determinstic by setting a lock.  Probably UNIX has the luxury of being
able to tell the process handling a signal to stop working for a second
while it fiddles with masks.  We don't have any reliable way of doing


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]