Re: AW: accept() failed message on kde 1.1.2 ??

[egor duda <deo at logos-m dot ru>]

Hi!

Friday, 15 June, 2001 Ralf Habacker Ralf.Habacker@saght.tessag.com wrote:

>> > The only change we had in the accept() code between 1.1.8 and
>> > 1.3.2 was Egor's patch to support secure connections on AF_UNIX
>> > sockets. Is it possible that the security check fails under
>> > some circumstances here?

you can apply this to cygwin sources, rebuild cygwin1.dll and see if
it helps. this disables security checks i've added to AF_UNIX socket
code. if it helps, then there may be some problem with my code.

Index: net.cc
===================================================================
RCS file: /cvs/src/src/winsup/cygwin/net.cc,v
retrieving revision 1.57
diff -u -p -1 -r1.57 net.cc
--- net.cc      2001/06/03 22:18:19     1.57
+++ net.cc      2001/06/15 09:27:01
@@ -765,2 +765,3 @@ cygwin_connect (int fd,
         }
+#if 0
       if (sock->get_addr_family () == AF_UNIX)
@@ -793,2 +794,3 @@ cygwin_connect (int fd,
         }
+#endif
     }
@@ -902,3 +904,3 @@ cygwin_accept (int fd, struct sockaddr *
       res = accept (sock->get_socket (), peer, len);  // can't use a blocking call inside a lock
-
+#if 0
       if (sock->get_addr_family () == AF_UNIX)
@@ -934,3 +936,3 @@ cygwin_accept (int fd, struct sockaddr *
         }
-
+#endif
       SetResourceLock (LOCK_FD_LIST, WRITE_LOCK|READ_LOCK, "accept");

RH> I will try, but the difficulty is that on stracing the error does not
RH> occure. I will analyze it in more detail ...

RH> an example

[...]

RH> Done anyone have an idea ?

unfortunately, i don't have kde installed so i can't get any
meaningful explanation of all those debug outputs. i suspect, however,
that this accept()-involved code could be somehow reduced to small
testcase.

you have already sent some bits to cygwin mailing list, and, if i
understand them correctly, it has some weird sequences of calls such
as 'conenect() to socket and then accept() on it without closing it
beforehand'. is that so? if you'll be able to extract socket-related
code that fails into reasonably small and self-contained program, i'll
be able to debug it.

Egor.            mailto:deo@logos-m.ru ICQ 5165414 FidoNet 2:5020/496.19