poll fixes infor for gnulib

repo_notes/gnulib_README.md

@@ -13,4 +13,5 @@ This is the largest concentration of changes as the fixes here can fix a whole h
 - [Minor FTS compile fix for windows](https://github.com/mitchcapper/gnulib/compare/master...ours_fts_minor_fixes?w=1)
 - [minor sysstat timespec fix](https://github.com/mitchcapper/gnulib/compare/master...ours_fix_sysstat_timespec_disable_bug?w=1) - gnulib polyfiller has an issue in Windows where it wouldn't be declared even when trying to define the replacement.
 - [build-aux improvements](https://github.com/mitchcapper/gnulib/compare/master...ours_build_aux_handle_dot_a_libs?w=1) -  For these wrappers around the compiler (`ar-lib`/`compile`) expand the conversion support of flags/libs to search more places and naming defaults with better shared/static lib support.  Adds optional color output support as well to the scripts if the proper environmental variables are set.  See `helpers.sh:setup_build_env` the `$BLD_CONFIG_LOG_COLOR_HIGHLIGHT` for an example. Even when not compiling gnulib based tools we often use these build scripts as they do a great job at handling gcc/g++ to windows compiler conversions.
-- [TEST define for physmem.c rename](https://github.com/mitchcapper/gnulib/compare/master...ours_physmem_test_main_rename?w=1) - physmem.c uses `#ifdef DEBUG` to determine if it should run test code, this is problematic if you want to define DEBUG but not have it in test mode (which won't compile as part of lib).  To fix renamed the define to TEST to match what other files use.
+- [TEST define for physmem.c rename](https://github.com/mitchcapper/gnulib/compare/master...ours_physmem_test_main_rename?w=1) - physmem.c uses `#ifdef DEBUG` to determine if it should run test code, this is problematic if you want to define DEBUG but not have it in test mode (which won't compile as part of lib).  To fix renamed the define to TEST to match what other files use.
+- [poll fixes](https://github.com/mitchcapper/gnulib/compare/master...ours_win32_poll_fixes?w=1) - gnulib has a poll() polyfill for windows, it seems to have some threading issues, and may return earlier than the timeout even though no data is ready.  To avoid breaking existing functionality for the timeout fix you must use poll_retry.