close to 1.3.4

Christopher Faylor cgf@redhat.com
Mon Oct 8 08:44:00 GMT 2001


On Mon, Oct 08, 2001 at 12:13:32PM +0400, egor duda wrote:
>Hi!
>
>Saturday, 06 October, 2001 Christopher Faylor cgf@redhat.com wrote:
>
>CF> On Fri, Oct 05, 2001 at 05:54:04PM -0400, Christopher Faylor wrote:
>>>On Fri, Oct 05, 2001 at 05:54:51PM +0400, egor duda wrote:
>>>>Wednesday, 03 October, 2001 Christopher Faylor cgf@redhat.com wrote:
>>>>CF> I'm a little reluctant to release 1.3.4 without Corinna's stabilizing
>>>>CF> presence but if no one has any objections, I'll probably do this in
>>>>CF> the next couple of days.
>>>>
>>>>i get a bunch of "mark_closed: attempt to close protected handle" in
>>>>'expect' while running gas testsuite. i don't know if it indicates a
>>>>real problem, but i'm looking at it now.
>>>
>>>So far, I haven't been able to duplicate this.  I'm trying again with various
>>>CYGWIN= settings.
>
>CF> Ok, I duplicated it with CYGWIN=tty ntsec.
>
>oh, i should have told what options i'm using from the beginning. will
>certainly do next time.
>
>CF>  I think I've checked in a fix for the problem.
>
>i wonder if such things should be a bit more "intrusive" in debugging
>version, i.e. should we call api_fatal on errors in add_handle and/or
>mark closed? if we do, i'd be easier to write test-cases for such
>problems. it won't affect normal users, since releases are built with
>debugging code disabled.

Good idea.  I think you're right that debugging version should exit on such
errors.

In this case the error was in the tracking of open/closed handles -- the
debugging code was at fault.  It almost feels like this happens 50% of
the time.  In the other 50% the ProtectHandle stuff actually unearths valid
bugs.  So it is useful.

So, whether it is a bug in the debugging code or an actual handle botch,
it would be good to have some mechanism for making these errors fatal.
So, I guess that changing the warning to an error (change system_printf
to api_fatal) in mark closed would be a good thing.

cgf



More information about the Cygwin-developers mailing list