[PATCH v7 1/1] Cygwin: pty: add pseudo console support.

[Takashi Yano]

Hi Corinna, Mark and Thomas,

On Sat, 17 Aug 2019 11:10:53 -0700 (PDT)
Mark Geisert wrote:
> On Sat, 17 Aug 2019, Corinna Vinschen wrote:
> > On Aug 16 01:09, Takashi Yano wrote:
> > It looks like there's some timing problem in terms of stderr output.  I
> > just made up the example for the Cygwin ML discussion about the mcmodel.
> > When I start the code built with -mcmodel=small I expected the output:
> >
> > Cygwin runtime failure: /home/corinna/dll/main.exe: Invalid relocation.
> > Offset 0xfffffffd80348989 at address 0x40000103b doesn't fit into 32 bits
> 
> That particular error message doesn't go out on Cygwin's stderr.  It's 
> done by a Windows WriteFile() on STD_ERROR_HANDLE followed by a 
> FlushFileBuffers() on same.
> 
> Not sure that it matters, but perhaps it does.

Mark, you are right.

On Thu, 15 Aug 2019 20:22:35 +0200
Thomas Wolff wrote:
> >> The same program, compiled with cygwin gcc, fails in mintty:
> >> output is skipped (like without the patch)
> >> ↯input stalls, kill -9 fails.
> >> Without conpty patch v6, both output and input are skipped.
> >> Note that winpty properly handles this case.
> > This problem is hard to be fixed in the current implementation...
> Do you have an idea why it could make a difference whether the program 
> is compiled with gcc or x86...mingw-gcc,
> even if only w32 API functions are called?

This is also the same problem.

The reason was already noted in:
https://cygwin.com/ml/cygwin-developers/2019-04/msg00042.html

Now I am under consideration on this problem.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>