Re: [PATCH 2/3] getdtablesize: Extend RLIMIT_NOFILE fallback case to Android

[Kevin Cernekee]

On Fri, Feb 20, 2015 at 11:00 AM, Paul Eggert <address@hidden> wrote:
> Kevin Cernekee wrote:
>
>> However, I don't see how any of the m4 changes address the problem
>> that is causing the test cases to fail: lib/getdtablesize.c still
>> wants to rely on sysconf(_SC_OPEN_MAX) on Android, and that call just
>> blindly returns 256 on all released versions of the OS.  This is why
>> my earlier submission changed it to use getrlimit() instead.
>
>
> Thanks for explaining.  How about if we go even further than that patch, and
> avoid the use of _SC_OPEN_MAX everywhere?  I applied the attached two
> patches; I think only the second one matters to you.  Do they fix the
> problem?

Unfortunately I don't think we define RLIM_SAVED_* on Android:

        arm-linux-androideabi-gcc -DHAVE_CONFIG_H -I. -I..
-DGNULIB_STRICT_CHECKING=1   -pie -fPIE -MT getdtablesize.o -MD -MP
-MF $depbase.Tpo -c -o getdtablesize.o getdtablesize.c &&\
        mv -f $depbase.Tpo $depbase.Po
getdtablesize.c: In function 'getdtablesize':
getdtablesize.c:107:26: error: 'RLIM_SAVED_CUR' undeclared (first use
in this function)
       && lim.rlim_cur != RLIM_SAVED_CUR
                          ^
getdtablesize.c:107:26: note: each undeclared identifier is reported
only once for each function it appears in
getdtablesize.c:108:26: error: 'RLIM_SAVED_MAX' undeclared (first use
in this function)
       && lim.rlim_cur != RLIM_SAVED_MAX)
                          ^
make[4]: *** [getdtablesize.o] Error 1
make[4]: Leaving directory `/shared/testdir/gllib'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/shared/testdir/gllib'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/shared/testdir/gllib'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/shared/testdir'
make: *** [all] Error 2


So maybe we need to add something like:

#ifndef RLIM_SAVED_CUR
#define RLIM_SAVED_CUR RLIM_INFINITY
#endif

Based on comments in tests/test-sys_resource.c it looks like somebody
may have started a <sys/resource.h> replacement, but there is more
work to be done.