[NaCl SDK] nacl_io: Plumb through {,f}utime{,s}

[NaCl SDK] nacl_io: Plumb through {,f}utime{,s}

* Switched KernelProxy to use utimens/futimens.
* kernel_intercept.* forwards utime/utimes/futimes -> {,f}utimens.
* Add Node::Futimens
* Override FuseFsNode::Futimens
* Add newlib headers to define utime/utimes/futimes
* Glibc uses at_stim as timespec instead, add define for st_atimensec for
  compatibility with newlib.

BUG=413412
R=sbc@chromium.org

Committed: https://crrev.com/3806e82b48f9695f939db28621f9a2009bbd2a33
Cr-Commit-Position: refs/heads/master@{#294678}

[binji, 2014-09-12 21:22:27]

binji@chromium.org changed reviewers:
+ bradnelson@chromium.org

[binji, 2014-09-12 21:22:27]

+bradnelson

[Sam Clegg, 2014-09-12 21:36:27]

great! lgtm

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/fuse.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/fuse.h:202: int (*utimens)(const char*,
const struct timespec tv[2]);
Why did you move this up?

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/include/sys/time.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/include/sys/time.h:15: int futimes(int
fd, const struct timeval times[2]);
Add TODO here to remove once newlib headers get these decls?

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/include/utime.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/include/utime.h:14: 
Add TODO here to remove once newlib headers get these decls?

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/kernel_proxy.cc (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/kernel_proxy.cc:885: int result =
futimens(fd, times);
Add explicit KernelProxy:: here to avoid confusion with C global functions with
the same name?

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/node.cc (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/node.cc:180: return EINVAL;
EACCES?  Or ENOSYS?   I guess we should be consistent with other stubs in this
file though.

[binji, 2014-09-12 21:50:08]

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/fuse.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/fuse.h:202: int (*utimens)(const char*,
const struct timespec tv[2]);
On 2014/09/12 21:36:26, Sam Clegg wrote:
> Why did you move this up?

I moved it because all functions below are not implemented. Though now that I
think of it, moving it could break users who have stubs for all of the functions
(even unimplemented ones).

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/include/sys/time.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/include/sys/time.h:15: int futimes(int
fd, const struct timeval times[2]);
On 2014/09/12 21:36:26, Sam Clegg wrote:
> Add TODO here to remove once newlib headers get these decls?

Done.

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/include/utime.h (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/include/utime.h:14: 
On 2014/09/12 21:36:27, Sam Clegg wrote:
> Add TODO here to remove once newlib headers get these decls?

Done.

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/kernel_proxy.cc (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/kernel_proxy.cc:885: int result =
futimens(fd, times);
On 2014/09/12 21:36:27, Sam Clegg wrote:
> Add explicit KernelProxy:: here to avoid confusion with C global functions
with
> the same name?  

We don't do that for open, close, stat, fstat, etc. Maybe we should go through
and do a pass in another CL?

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
File native_client_sdk/src/libraries/nacl_io/node.cc (right):

https://codereview.chromium.org/565343002/diff/1/native_client_sdk/src/librar...
native_client_sdk/src/libraries/nacl_io/node.cc:180: return EINVAL;
On 2014/09/12 21:36:27, Sam Clegg wrote:
> EACCES?  Or ENOSYS?   I guess we should be consistent with other stubs in this
> file though.

Yeah, I was just being consistent. But ENOSYS is probably better...

[Sam Clegg, 2014-09-12 21:51:52]

lgtm

[commit-bot: I haz the power, 2014-09-12 22:18:40]

Patchset 2 (id:??) landed as
https://crrev.com/3806e82b48f9695f939db28621f9a2009bbd2a33
Cr-Commit-Position: refs/heads/master@{#294678}

[binji, 2014-09-12 22:18:40]

Committed patchset #2 (id:20001) manually as 3806e82 (presubmit successful).

[Sam Clegg, 2014-09-12 23:33:04]

On 2014/09/12 22:18:40, binji wrote:
> Committed patchset #2 (id:20001) manually as 3806e82 (presubmit successful).

Oh.. I think we might run into issues when this ends up in naclports build
because of ports/glibc-compat/include/sys/utime.h.

[bradn, 2014-09-13 21:06:32]

bradnelson@google.com changed reviewers:
+ bradnelson@google.com

[bradn, 2014-09-13 21:06:33]

I think returning EINVAL is causing problems with a bunch of the ports.
It throws off all the tar implementations for instance.