Re: Problem with sys_select on Cygwin

[Ken Brown]

On 3/21/2016 1:32 PM, Paul Eggert wrote:
> Eric Blake wrote:
>> This feels a bit too broad - it assumes that everyone is compiling with
>> ONLY the latest newlib headers; remember, the cygwin header changes have
>> NOT been released yet, but are in beta testing to see what else needs to
>> change at the same time as the cygwin release.
> 
> As far as I know the current version should work OK with older Cygwin, 
> as the affected #includes are needed only for non-Cygwin systems.

I think that's right.

There's still the question raised by Corinna (on the Cygwin mailing list) as to 
whether we should be testing for __NEWLIB__ rather than __CYGWIN__.  I'm 
inclined to say yes, just for the sake of clarity, since the headers in 
question come from newlib.  In other words, I would suggest the following, if 
Eric agrees:

--- a/lib/sys_select.in.h
+++ b/lib/sys_select.in.h
@@ -82,8 +82,8 @@
    Also, Mac OS X, AIX, HP-UX, IRIX, Solaris, Interix declare select()
    in <sys/time.h>.
    But avoid namespace pollution on glibc systems and "unknown type
-   name" problems on Cygwin.  */
-# if !(defined __GLIBC__ || defined __CYGWIN__)
+   name" problems on newlib systems.  */
+# if !(defined __GLIBC__ || defined __NEWLIB__)
 #  include <sys/time.h>
 # endif

@@ -102,10 +102,10 @@

 /* Get definition of 'sigset_t'.
    But avoid namespace pollution on glibc systems and "unknown type
-   name" problems on Cygwin.
+   name" problems on newlib systems..
    Do this after the include_next (for the sake of OpenBSD 5.0) but before
    the split double-inclusion guard (for the sake of Solaris).  */
-#if !((defined __GLIBC__ || defined __CYGWIN__) && !defined __UCLIBC__)
+#if !((defined __GLIBC__ || defined __NEWLIB__) && !defined __UCLIBC__)
 # include <signal.h>
 #endif


Ken