plumb glibc clock_get interfaces to irt

plumb glibc clock_get interfaces to irt

cargo-culted changes to add sysdeps/nacl/ files to librt's
clock_gettime/clock_getres implementation.  seems to work.

R=mcgrathr@google.com
BUG= http://code.google.com/p/nativeclient/issues/detail?id=2477
TEST= in separate CL that cannot be submitted w/o TC DEPS roll

Committed: https://git.chromium.org/gitweb?p=native_client/nacl-glibc.git;a=commit;h=d5d66ea

[Roland McGrath, 2012-02-14 00:15:44]

This is probably fine as is, so provisional LGTM.
Whitespace trivia below.

One more substantive note is wrt CLOCK_THREAD_CPUTIME et al.  The glibc
code has paths that answer those using rdtsc when the OS doesn't provide
any mechanism.  That gives lousy quality and is not robust to scheduling
issues and so on.  But some people think it's better than nothing.  I'm not
entirely sure whether nacl-glibc's clock_gettime were using those fallback
paths before.  If they were, then this could be a marginal regression in
that previously we'd give lousy answers (but that might in some
circumstances be almost useful) and now we'll fail with EINVAL or suchlike.
So it merits doing a quick empirical test to see if clock_gettime on
CLOCK_THREAD_CPUTIME previously worked or not.

If we wanted to support that path, then the
sysdeps/nacl/clock_{gettime,getres}.c files should #define the
SYSDEP_GETTIME macro to handle what the IRT calls handle, and
then #include <sysdeps/unix/clock_gettime.c> to get the CPUTIME cases.

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
File sysdeps/nacl/clock_getres.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:7: 
excess blank lines

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:12: if (result != 0) {
glibc style puts { on a new line here, though style issues have been pretty
random in the nacl-glibc code so it doesn't really matter.

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:13: __set_errno(result);
space before paren here

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_gett...
File sysdeps/nacl/clock_gettime.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_gett...
sysdeps/nacl/clock_gettime.c:7: 
excess blank line here.

[bsy, 2012-02-15 00:01:13]

your suspicions are on the nose.  the current glibc TC does allow
CLOCK_THREAD_CPUTIME_ID / CLOCK_PROCESS_CPUTIME_ID uses, so i'm going to tweak
this as you suggests.  will send a ptal later.

[bsy, 2012-02-15 18:35:59]

PTAL

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
File sysdeps/nacl/clock_getres.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:7: 
On 2012/02/14 00:15:45, Roland McGrath wrote:
> excess blank lines

Done.

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:12: if (result != 0) {
On 2012/02/14 00:15:45, Roland McGrath wrote:
> glibc style puts { on a new line here, though style issues have been pretty
> random in the nacl-glibc code so it doesn't really matter.

Done.

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_getr...
sysdeps/nacl/clock_getres.c:13: __set_errno(result);
On 2012/02/14 00:15:45, Roland McGrath wrote:
> space before paren here

Done.

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_gett...
File sysdeps/nacl/clock_gettime.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/1/sysdeps/nacl/clock_gett...
sysdeps/nacl/clock_gettime.c:7: 
On 2012/02/14 00:15:45, Roland McGrath wrote:
> excess blank line here.

Done.

[Roland McGrath, 2012-02-15 18:45:25]

LGTM with changes below and happy testing.

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_getres.c
File sysdeps/nacl/clock_getres.c (right):

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_getres.c#...
sysdeps/nacl/clock_getres.c:10: int result = __nacl_irt_clock_getres (clock_id,
res); \
GNU style says this should be indented only two spaces from the brace.

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_getres.c#...
sysdeps/nacl/clock_getres.c:12: {                                               
     \
GNU style says this whole block is indented two spaces from the if (and likewise
for the else).

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_getres.c#...
sysdeps/nacl/clock_getres.c:17: {                                               
     \
GNU style avoids superfluous braces for single-line cases.

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_gettime.c
File sysdeps/nacl/clock_gettime.c (right):

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_gettime.c...
sysdeps/nacl/clock_gettime.c:12: {                                              
      \
Indent block.

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_gettime.c...
sysdeps/nacl/clock_gettime.c:17: {                                              
      \
Drop braces.

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_nanosleep.c
File sysdeps/nacl/clock_nanosleep.c (right):

http://codereview.chromium.org/9389024/diff/5001/sysdeps/nacl/clock_nanosleep...
sysdeps/nacl/clock_nanosleep.c:1: #include <rt/clock_nanosleep.c>
Use sysdeps/unix/clock_nanosleep.c instead.
It supports the CLOCK_REALTIME case, which is the one case we can support (by
calling nanosleep).

[bsy, 2012-02-15 19:33:29]

done.  thx.

will rebuild toolchain & test before committing.

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
File sysdeps/nacl/clock_getres.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
sysdeps/nacl/clock_getres.c:10: int result = __nacl_irt_clock_getres (clock_id,
res); \
On 2012/02/15 18:45:26, Roland McGrath wrote:
> GNU style says this should be indented only two spaces from the brace.

Done.  (well, i indented the brace further.)

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
sysdeps/nacl/clock_getres.c:12: {                                               
     \
On 2012/02/15 18:45:26, Roland McGrath wrote:
> GNU style says this whole block is indented two spaces from the if (and
likewise
> for the else).

Done.

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
sysdeps/nacl/clock_getres.c:17: {                                               
     \
On 2012/02/15 18:45:26, Roland McGrath wrote:
> GNU style avoids superfluous braces for single-line cases.

Done.

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
File sysdeps/nacl/clock_gettime.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
sysdeps/nacl/clock_gettime.c:12: {                                              
      \
On 2012/02/15 18:45:26, Roland McGrath wrote:
> Indent block.

Done.

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_g...
sysdeps/nacl/clock_gettime.c:17: {                                              
      \
On 2012/02/15 18:45:26, Roland McGrath wrote:
> Drop braces.

Done.

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_n...
File sysdeps/nacl/clock_nanosleep.c (right):

https://chromiumcodereview.appspot.com/9389024/diff/5001/sysdeps/nacl/clock_n...
sysdeps/nacl/clock_nanosleep.c:1: #include <rt/clock_nanosleep.c>
On 2012/02/15 18:45:26, Roland McGrath wrote:
> Use sysdeps/unix/clock_nanosleep.c instead.
> It supports the CLOCK_REALTIME case, which is the one case we can support (by
> calling nanosleep).

Done.