KO Myung-Hun wrote: > > If you set gl_cv_func_popen_works to 'no', it will activate the workaround > > in lib/popen.c. Does this workaround work? Does 'test-[p]open' succeed? > > Do you mean 'test-popen'? Then it works.
Good. Then lib/popen.c does not need to be extended. > > If yes, we'll also need an edit to doc/posix-functions/popen.texi. > > > > Part of 'Portability problems fixed by Gnulib:'? Yes. Where it lists Cygwin as relevant platform, please list also OS/2 with kLIBC. > >> else > >> + dnl On OS/2 kLIBC, the following test hangs on. So set to no here. > >> + case "$host_os" in os2*) gl_cv_func_popen_works=no;; esac > >> AC_CACHE_CHECK([whether popen works with closed stdin], > >> [gl_cv_func_popen_works], > >> [ > >> > > > > Can you put the 'gl_cv_func_popen_works=no' into the AC_CACHE_CHECK? > > And produce the value 'guessing no', since the test is not being run. > > Rationale: We want 'configure' to print > > checking whether popen works with closed stdin... guessing no > > not > > checking whether popen works with closed stdin... (cached) no > > If the test is run, 'configure' is blocked until pressing Ctrl-C or > Ctrl-Break. Or do you mean modification of the test source codes to > return an error if OS/2? I mean, in the third argument of AC_CACHE_CHECK, which is a set of shell statements, add the case "$host_os" in os2*) there, before the test program gets created. Similar to the first AC_CACHE_CHECK in m4/fenv-environment.m4. Bruno
