Thanks for very the detailed bug report. This should be fixed in
the latest snapshot (when it becomes available).
The plethora of details and your actual honest-to-god test case helped
a great deal in tracking down this problem. If every bug report was
like this one, then fixing bugs would almost be a pleasure rather than
a chore.
Thanks again,
cgf
On Mon, Oct 01, 2001 at 11:06:52PM +0900, Shun-ichi GOTO wrote:
>Hi, cygwin users/developers
>
>## I'm not ML member, please add Cc: [EMAIL PROTECTED] for this thread.
>
>I'm using latest snapshot and encounted to problem around select() for
>PIPE input.
>
>The problem is:
> select() onto PIPE input returns immediately as "something exist"
> on command prompt (cmd.exe) and 4nt.
>
>Try attached test program "pipe-select.c" and examine following two
>case with some version of DLLs.
>
>(1) pipe-select
>
>(2) cat - | pipe-select
>
>With good DLL, (1) waits keyboard input by select() and wait...
>When I hit any key, program goes read() and until EOL is start. And repeat.
>(2) waits pipe input by select() and wait.
>When I enter some chars and ENTER, then program goes read().
>Both are good behavior.
>
>With bad DLL, (1) is same to good DLL. But (2) goes read() immediately
>although no data is available in PIPE.
>
>This problem means asyncronous reading using select() is not work if
>some input handle is PIPE. For example, OpenSSH executed from NTEmacs is
>not work. On this case, ssh.exe goes reading from stdin (PIPE) although
>PIPE is empty and blocks processing input from network.
>
>Described above is phenomena on non cygwin shell (cmd.exe and 4nt.exe).
>On bash, it seems working good. But once hit CTRL-C to stop select(),
>it hangs. Program cannot terminate select'ing.
>It's another problem.
>
>
>I checked tests above on command prompt and 4nt.exe on Windows XP and
>Windows 2000 with some cygwin1.dll (ver 1.3.3, snap-20010925,
>snap-20010927, snap-20010929, snap-20011001) and get results bellow.
>
>cygwin1.dll result
>-----------------------
>1.3.3 good
>snap-20010925 BAD
>snap-20010927 BAD
>snap-20010929 good
>snap-20011001 BAD
>
>
>
>/* pipe-select.c -- test pipe input select()'ing bug. */
>
>#include <stdio.h>
>#include <sys/types.h>
>#include <sys/select.h>
>
>int
>main(int argc, char **argv)
>{
> int ret;
> fd_set ifds;
>
> if ( isatty(0) ) {
> puts("stdin is tty");
> } else {
> puts("stdin is NOT tty");
> }
>
> FD_ZERO(&ifds);
> FD_SET(0, &ifds);
> printf("selecting..."); fflush(stdout);
> ret = select( 1, &ifds, NULL, NULL, NULL);
> printf("ret=%d, ifds=0x%x\n", ret, *(long*)&ifds);
> if ( FD_ISSET(0,&ifds) ) {
> char c;;
> printf("reading: ");fflush(stdout);
> read(0, &c, 1);
> printf("=> '%c'\n", c);
> }
> return 0;
>}
>// Local Variables:
>// compile-command: "gcc pipe-select.c -o pipe-select"
>// End:
>
>
>--- Regards,
> Shun-ichi Goto <[EMAIL PROTECTED]>
> R&D Group, TAIYO Corp., Tokyo, JAPAN
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
