Re: cygrunsrv + sshd + rsync = 20 times too slow -- throttled?

[Ken Brown <kbrown@cornell.edu>]

On 9/20/2021 5:27 PM, Takashi Yano wrote:
> Hi Ken,
> 
> Thanks for reviewing the code.
> 
> On Mon, 20 Sep 2021 17:09:48 -0400
> Ken Brown wrote:
>> On 9/20/2021 3:14 PM, Ken Brown wrote:
>>> +      if (wcscmp (name, ntfn->Name.Buffer) == 0)
>>> +	{
>>> +	  query_hdl_proc = proc;
>>> +	  query_hdl_value = (HANDLE)(intptr_t) shi->Handles[i].HandleValue;
>>> +	  qh = h;
>>> +	  break;
>>> +	}
>>> +close_handle:
>>> +      CloseHandle (h);
>>> +close_proc:
>>> +      CloseHandle (proc);
>>
>> Doesn't this mean that query_hdl_proc is not a valid handle any more?  So the
>> attempt to reduce overhead at the beginning of tentative_query_hdl() will never
>> work.
> 
> When the handle which name matches the pipe name is found, 'break'
> is used in the 'for' loop, so the CloseHandle(proc) is not called.
> Therefore, query_hdl_proc still has valid handle after 'break'.

Right, I realized this myself.  I guess you wrote your reply before seeing my 
later message in which I said "I see my stupid mistake".
> 
>> Other than that, I still think it looks great.  But it's complicated enough that
>> I think Corinna should review it too when she returns.  (It also uses Windows
>> functions that I have no experience with.)  Nevertheless, I'm tempted to push it
>> so that it can get testing, even if Corinna changes it or reverts it later.

I'll push it tomorrow unless you find something that needs to be changed before 
then.

Ken