Hi Corinna, On Tue, 4 Jul 2023 16:36:26 +0200 Corinna Vinschen wrote: > On Jul 4 19:03, Takashi Yano wrote: > > This old kludge code assigns fhandler_console for /dev/tty even > > if the CTTY is not a console when stat() has been called. Due to > > this, the problem reported in > > https://cygwin.com/pipermail/cygwin/2023-June/253888.html > > occurs after the commit 3721a756b0d8 ("Cygwin: console: Make the > > console accessible from other terminals."). > > > > This patch fixes the issue by dropping the old kludge code. > > > > Though the exact reason why the kludge code was necessary is not > > clear enough, this kluge code has no longer seemed to be necessary > ^^^^^^^^^^^^^^^^^^^^ > I'm not a native speaker myself, but > > no longer seems > > might be better here. > > Anyway, this is GTG.
I think I understand correctly the concept of cnew_no_ctor macro in dtable.cc now. cnew_no_ctor calls fhandler_console(void *) instead of fhandler_console(fh_devices) to omits initialization of instance for stat() call. This might make stat() slightly faster. Based on this understanding, I would like to withdraw the previous patch, and propose new patch series. Could you please review the patch seriese? -- Takashi Yano <takashi.y...@nifty.ne.jp>