[PATCH] Cygwin: pipe: Restore blocking mode of read pipe on close()

[Corinna Vinschen]

Hi Takashi,

I think you should push your original patch to the 3.5 branch for now,
and we test the big change in the main branch first.


Thanks,
Corinna


On Sep  5 22:18, Takashi Yano wrote:
> Hi Corinna, Ken and Johannes
> 
> On Mon, 2 Sep 2024 23:33:13 +0900
> Takashi Yano wrote:
> > On Mon, 2 Sep 2024 22:50:45 +0900
> > Takashi Yano wrote:
> > > On Mon, 2 Sep 2024 14:48:35 +0200 (CEST)
> > > Johannes Schindelin wrote:
> > > > I have tested it and the symptom is addressed.
> > > 
> > > Thanks for testing.
> > > 
> > > > I do have to wonder whether it is intentional that calling
> > > > `set_pipe_non_blocking(false)` followed by `set_pipe_non_blocking(true)`
> > > > on an originally-non-blocking pipe will "restore" it to blocking mode,
> > > > though.
> > > 
> > > I'm not sure how such symptom occurs.
> > > 
> > > Calling `set_pipe_non_blocking(true)` on an originally-non-blocking pipe
> > > will set `was_blocking_read_pipe` to false.
> > > 
> > > Furthermore, regardless of the value of `was_blocking_read_pipe`, calling
> > > set_pipe_non_blocking(false) always sets the pipe blocking mode. It is not
> > > due to "restore" logic.
> > 
> > Ah...
> > However, if a cygwin app executes non-cygwin app and the non-cygwin app exits,
> > the read pipe remains on blocking mode. Then the cygwin app cannot handle
> > signal on blocking read() after that.
> > 
> > Let me consider...
> 
> I have submit a new patch for these issues.
> With this pathch the strategy of pipe read/write for blocking/non-blocking
> is re-designed. To eliminate toggling the pipe blocking mode, cygwin pipe
> is always blocking mode basically, and non-blocking mode is sumulated by
> raw_read() and raw_write().
> 
> The new patch will be posted shortly with subject:
> [PATCH] Cygwin: pipe: Switch pipe mode to blocking mode by defaut
> 
> What do you think of this idea?
> 
> -- 
> Takashi Yano <takashi.yano@nifty.ne.jp>