From: Christopher Faylor Subject: Re: services not starting with 20060104
snapshot
Date: Thu, 5 Jan 2006 01:05:45 -0500
On Thu, Jan 05, 2006 at 12:42:56AM -0500, Igor Peshansky wrote:
>On Wed, 4 Jan 2006, Christopher Faylor wrote:
>
>> On Wed, Jan 04, 2006 at 11:22:13PM -0500, Igor Peshansky wrote:
>> >On Wed, 4 Jan 2006, Christopher Faylor wrote:
>> >>On Wed, Jan 04, 2006 at 08:45:15PM -0500, Igor Peshansky wrote:
>> >>>On Wed, 4 Jan 2006, Christopher Faylor wrote:
>> >>>
>> >>>>On Wed, Jan 04, 2006 at 06:38:26PM -0500, Igor Peshansky wrote:
>> >>>>>Frankly, I'm not sure what the right fix would be in this case, or
even
>> >>>>>how to debug this... Any ideas on how I can simulate service
startup
>> >>>>>on the command line (to allow popping up gdb) would be greatly
>> >>>>>appreciated.
>> >>>>
>> >>>>Does the change I just checked into CVS cause any difference in
>> >>>>behavior? If not, could you also try changing the
CreateWindowStation
>> >>>>argument &sec_all_nih to NULL?
>> >>>
>> >>>Nope, sorry to say, neither helped. In both cases I got the same
error
>> >>>1053 when starting the service. The Windows error log messages are
>> >>>identical as well.
>> >>
>> >>How about the latest CVS? This will probably revert to the "ssh
>> >>localhost date" failing scenario, I assume.
>> >
>> >Sigh. Exactly. The service starts, though...
>> >
>> >I wonder if the service startup problem has to do with the fact that
>> >spawn_guts() calls GetProcessWindowStation() and isn't getting the one
>> >you create in fhandler_console::need_invisible()... Hmm, no, you call
>> >SetProcessWindowStation() there too... I can't think of a way for me
>> >to check whether it succeeds, though.
>>
>> Can you try inverting the (now) two CreateWindowStation calls in
>> need_invisible so that the "CygwinInvisible" case is called first and
>> the NULL case is called second?
>
>Nope, didn't work. The service doesn't start in this case.
>
>> And, if that doesn't work, then try commenting out the
>> CloseWindowStation.
>
>I wasn't sure whether you meant commenting it out after the swap or
>before, so I tried both. It didn't affect the behavior in either case
>(service not started with the swap; no output from ssh without the swap).
Any and all combinations are welcome at this point. There seem to be an
infinite number of straws.
Next straw would be to put a call to init_console_handler(TRUE); just
before
the call to CloseWindowStation:
if (horig && h && h != horig && SetProcessWindowStation (horig))
{
set_console_handler_true (TRUE);
CloseWindowStation (h);
}
>> Also, what version of XP are you running? Pro? Is it up-to-date with
>> all service packs and updates?
>
>I'm running XP Pro SP1, with most updates applied (except KB835409 and
>KB910437, which are pending).
Maybe that is a clue, as Yitzchak suggested. I'd missed that in the
cygcheck
output. I'm running SP2 and I believe Corinna is also.