On Mar 19 18:05, Corinna Vinschen wrote: > On Mar 19 12:46, Ken Brown wrote: > > On 3/19/2014 12:09 PM, Corinna Vinschen wrote: > > >On Mar 19 11:37, Ken Brown wrote: > > >>Here's some more weirdness involving the recent changes to exception > > >>handling on 64-bit Cygwin. > > >> > > >>I have an ordinary user kbrown and an administrator user > > >>kbrown-admin. I use the following command to simulate "su": > > >> > > >> ssh kbrown-admin@$(hostname) . > > >> > > >>With the current (2014-03-18) x86_64 snapshot, this doesn't work if > > >>I try to use public key authorization. I simply get a message > > >> > > >> Connection closed by fe80::81db:bf3a:6434:6015%11 > > >> > > >>and I'm not even prompted for my passphrase [*]. But if I move my > > >>id_rsa and id_rsa.pub out of the way, I'm prompted for a password, > > >>and I can login successfully. > > >> > > >>Reverting to the 2014-03-17 snapshot fixes the problem. > > > > > >I just checked in a change. Please give it a try. If this doesn't work > > >either, I don't know how to proceed. > > > > This fixes the public key authorization problem, but the timeout > > during the starting of gtk applications is back. The > > "window-default" program from the other thread is still a test case > > for that problem. > > Ok, I give up. I have no idea anymore how to change that.
I have a big problem here. I reverted the entire exception handling to 1.7.28 state, and the "window-default" application still hangs for me and generates the WARNING **: Error retrieving accessibility bus address: org.freedesktop.DBus.Error.NoReply: Did not receive a reply [etc] message before the window opens. Then I rebuild the same code, without any changes, just with -g -O2 instead of just -g, and the window-default application didn't hang. If I build everything with -g and only the files which include exception.h with -O2, I would have expected that it works, but not so. It still hangs. I tried every variation of execption handler installation and the kitchen sink, but whatever change I try, the only variation which works for window-default is the 1.7.28 version compiled with -O2. Which is not good given that the exception handler is called twice. Every change to make sure the vectored continue handler is only called in the debugger result in window-default hanging again. I just can't find a stable state which does not fail one way or the other :( We need to know what happens there, but the various dbus-FOO processes started at once are a bit overwhelming. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Maintainer cygwin AT cygwin DOT com Red Hat
pgpbGcBbdDjoh.pgp
Description: PGP signature
