[#5219] Segmentation fault in timeout.rb — Michel Pastor <K@...>

Hi,

18 messages 2005/06/16
[#5220] Re: Segmentation fault in timeout.rb — Eric Hodel <drbrain@...7.net> 2005/06/16

[#5221] Re: Segmentation fault in timeout.rb — Michel Pastor <K@...> 2005/06/16

On Fri, 17 Jun 2005 05:03:18 +0900

[#5223] Re: Segmentation fault in timeout.rb — nobu.nokada@... 2005/06/17

Hi,

[#5296] Subversion — Shugo Maeda <shugo@...>

Hi,

64 messages 2005/06/30
[#5297] Re: Subversion — Curt Hibbs <curt@...> 2005/06/30

Shugo Maeda wrote:

[#5298] Re: Subversion — Nikolai Weibull <mailing-lists.ruby-core@...> 2005/06/30

Curt Hibbs wrote:

[#5301] Re: Subversion — Austin Ziegler <halostatue@...> 2005/06/30

On 6/30/05, Nikolai Weibull

[#5304] Re: Subversion — Nikolai Weibull <mailing-lists.ruby-core@...> 2005/06/30

Austin Ziegler wrote:

[#5305] Re: Subversion — Austin Ziegler <halostatue@...> 2005/06/30

On 6/30/05, Nikolai Weibull

[#5307] Re: Subversion — mathew <meta@...> 2005/06/30

Austin Ziegler wrote:

[#5308] Re: Subversion — Austin Ziegler <halostatue@...> 2005/06/30

On 6/30/05, mathew <meta@pobox.com> wrote:

[#5311] Re: Subversion — mathew <meta@...> 2005/07/01

Austin Ziegler wrote:

[#5323] Re: Subversion — Austin Ziegler <halostatue@...> 2005/07/01

On 7/1/05, mathew <meta@pobox.com> wrote:

[#5325] Re: Subversion — Nikolai Weibull <mailing-lists.ruby-core@...> 2005/07/01

Austin Ziegler wrote:

fcntls only partially supported on non-Unix platform

From: "BG - Ben Armstrong" <BArmstrong@...>
Date: 2005-06-03 16:28:11 UTC
List: ruby-core #5147
In ruby/lib/webrick/utils.rb there is this test:

      if defined?(Fcntl::FD_CLOEXEC)
        io.fcntl(Fcntl::FD_CLOEXEC, 1)
      end

Unfortunately, my platform, VMS, supports FD_CLOEXEC, so it is defined,
but it won't work in some cases, (presumably when called on a socket,
though I have not refined my failing case to know this for certain.)
Thus, the test succeeds but the actual fcntl, when issued, throws an
uncaught exception.

I'm not sure what to do in this general case of partially-supported
fcntls.  I suppose one kludge would be to just not define the fcntl, but
that unnecessarily makes the fcntl unavailable for the other cases.
Another possibility would be to fake it in io.fcntl, pretending the call
succeeded when it didn't, though that could have unforeseen and
unpleasant consequences, depending on the particular fcntl.  Or we could
simply error-trap every fcntl in ruby/lib to define the behaviour if the
fcntl is defined but throws an error.  This seems like the cleanest
solution, but it also sounds like a lot of work.

Ben


In This Thread

Prev Next