some bug fixes

Chris Faylor
Wed Feb 23 20:15:00 GMT 2000

On Fri, Feb 18, 2000 at 01:48:35PM -0000, Fifer, Eric wrote:
>These diffs are against the Feb-4 snapshot (more recents
>snapshots don't seem to be working for me) ...

Thanks for the patches.  Sorry it has taken so long to respond
but I've been on vacation (more or less).

I've forwarded your newlib change to the newlib group.

You're correct in finding that signals are not getting flushed after a
mask is released.  Thanks for tracking this down.  I've chosen to do this
using the sig_dispatch_pending () function which was designed for this.

I don't understand your select patch, so I'm not going to check it in.
The ChangeLog entry seems to indicate that it has something to do with
except_selected, but your patch doesn't enforce that.  Also, a "strcmp"
is not the way to find out what kind of device is in use.  You can compare
the device to the FH_PIPEW constant.

Thanks again for your patch submission.

FYI, in the future, if you submit patches (and I hope you do) please don't
provide a diff of the ChangeLog.  The convention is to just provide the
ChangeLog entries themselves.

Also, if you plan on contributing code you should fill out the
assignment form at .
It's an unfortunate legal requirement that we require this for our net


>+++ newlib/ChangeLog    Fri Feb 18 13:31:18 2000
>@@ -1,3 +1,9 @@
>+Fri Feb 18 13:25:59 2000  Eric Fifer <>
>+       * libc/stdio/findfp.c (__sinit): stdout should be fully
>+       buffered by default, code in makebuf.c (__smakebuf) already
>+       asserts line buffering on tty devices.
>+++ winsup/cygwin/ChangeLog     Fri Feb 18 13:25:02 2000
>@@ -1,3 +1,10 @@
>+Fri Feb 18 13:13:52 2000  Eric Fifer <>
>+       * (set_process_mask): flush pending signals
>+       freed by sigmask change.
>+       * (peek_pipe): ignore failure of PeekNamedPipe
>+       when except_selected on write end of a pipe.

More information about the Cygwin-developers mailing list