X-Recipient: archive-cygwin AT delorie DOT com
X-Spam-Check-By: sourceware.org
Date: Sun, 10 May 2009 20:27:39 -0400
From: Christopher Faylor <cgf-use-the-mailinglist-please AT cygwin DOT com>
To: cygwin AT cygwin DOT com
Subject: Re: How to detect a cygwin thread?
Message-ID: <20090511002739.GF25909@ednor.casa.cgf.cx>
Reply-To: cygwin AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
References: <9f8a01cd0905091706s6944a639m8da2f943212cc178 AT mail DOT gmail DOT com> <loom DOT 20090510T004554-316 AT post DOT gmane DOT org> <9f8a01cd0905100245m16838bb9w3c6e494d4a03a4cb AT mail DOT gmail DOT com> <loom DOT 20090510T184100-821 AT post DOT gmane DOT org> <20090510202132 DOT GB25909 AT ednor DOT casa DOT cgf DOT cx> <9f8a01cd0905101533i2902636aub172298be61599a5 AT mail DOT gmail DOT com> <20090510233629 DOT GC25909 AT ednor DOT casa DOT cgf DOT cx> <9f8a01cd0905101707x2b91f1edg33ae17ec82ab722e AT mail DOT gmail DOT com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <9f8a01cd0905101707x2b91f1edg33ae17ec82ab722e@mail.gmail.com>
User-Agent: Mutt/1.5.16 (2007-06-09)
Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
Precedence: bulk
List-Id: <cygwin.cygwin.com>
List-Unsubscribe: <mailto:cygwin-unsubscribe-archive-cygwin=delorie DOT com AT cygwin DOT com>
List-Subscribe: <mailto:cygwin-subscribe AT cygwin DOT com>
List-Archive: <http://sourceware.org/ml/cygwin/>
List-Post: <mailto:cygwin AT cygwin DOT com>
List-Help: <mailto:cygwin-help AT cygwin DOT com>, <http://sourceware.org/ml/#faqs>
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
Delivered-To: mailing list cygwin AT cygwin DOT com

On Mon, May 11, 2009 at 02:07:51AM +0200, Piotr Wyderski wrote:
>Christopher Faylor wrote:
>>Yes, that's the signal thread but I don't know why stopping it would
>>cause any special problems since, if the entire program is stopped, it
>>isn't going to be processing signals.
>
>I don't know the reason, just report the program's behaviour.  But to
>provide you more context: this code is a critical error handler.  All
>it has to do is:
>
>1.  immediately stop the world in a nice way, as this thread is all
>about.  When "sig" is suspended, no more points follow.
>
>2.  dump detailed stack traces + module list + the emergency stop
>reason description to the set of log files;
>
>3.  call std::abort in order to generate a core file for post-mortem
>analysis.

I wasn't asking for more context.  This isn't a bug in Cygwin.

I'm telling you that I can't think of a good reason for the suspension
of the signal thread to cause a hang other than the fact that Windows
probably doesn't like calling SuspendThread on a thread which is
blocking in ReadFile or something similar.

Your takeaway from this is:

1) I don't care about this.  Not suspending the "sig" thread fixes my
problem.

or

2) I'd better check to see if that is the case since my techniques to
"stop the world" will fail if any thread is calling ReadFile.

>> ?If that is the case, then you are going to see
>> problems any time a thread is doing blocking I/O.
>
>Why is that?  MSDN for SuspendThread() indicates no such issue.  There
>is an obvious warning about possible deadlock when the suspended thread
>has a synchronization object needed by the suspender, but it is not the
>case here.

There is nothing in the SuspendThread documentation to indicate that it
*won't* block either.

You might notice the hoops that are jumped through in the Cygwin signal
code to try to call SuspendThread only when it is known to be safe.  We
try hard not to call SuspendThread when the program is in the middle of
a Windows call.  We do that because we're trying to avoid random hangs
and SEGVs which we've noticed through the years when we have attempted
to use SuspendThread.

>BTW, is there a plan to support backtrace() and backtrace_symbols() (or
>something compatible)?

Nope.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/