Do fcntl(...F_FULLFSYNC...) on mac instead of fsync

Do fcntl(...F_FULLFSYNC...) on mac instead of fsync

From mac's fsync man page:

For applications that require tighter guarantees about the integrity of their
data, Mac OS X provides the F_FULLFSYNC fcntl.  The F_FULLFSYNC fcntl asks the
drive to flush all buffered data to permanent storage.  Applications, such as
databases, that require a strict ordering of writes should use F_FULLFSYNC to
ensure that their data is written in the order they expect.

Also, specifically check for a return value of -1 from fdatasync. mac's fcntl
only guarantees to return -1 on error and something other than -1 on success.
All other platforms agree to return -1 on error for fsync/fdatasync/_commit.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=221413

[dgrogan, 2013-09-04 23:25:51]

Josh, could you review this?

It can only hurt performance but looks like we are doing the wrong thing on mac.

[jsbell, 2013-09-04 23:59:34]

lgtm

https://codereview.chromium.org/23549017/diff/1/third_party/leveldatabase/env...
File third_party/leveldatabase/env_chromium.cc (right):

https://codereview.chromium.org/23549017/diff/1/third_party/leveldatabase/env...
third_party/leveldatabase/env_chromium.cc:72: return HANDLE_EINTR(fcntl(fildes,
F_FULLFSYNC, 0));
Can fcntl() produce an EINTR error on MacOS with F_FULLFSYNC? The man page
appears to indicate it will only return that if F_SETLKW is returned.

It should be harmless to use this wrapper if that's the case, though.

https://codereview.chromium.org/23549017/diff/1/third_party/leveldatabase/env...
third_party/leveldatabase/env_chromium.cc:515: if (fdatasync(fileno(file_)) ==
-1 && !error)
Nice catch.

[dgrogan, 2013-09-05 00:01:32]

https://codereview.chromium.org/23549017/diff/1/third_party/leveldatabase/env...
File third_party/leveldatabase/env_chromium.cc (right):

https://codereview.chromium.org/23549017/diff/1/third_party/leveldatabase/env...
third_party/leveldatabase/env_chromium.cc:72: return HANDLE_EINTR(fcntl(fildes,
F_FULLFSYNC, 0));
On 2013/09/04 23:59:34, jsbell wrote:
> Can fcntl() produce an EINTR error on MacOS with F_FULLFSYNC? The man page
> appears to indicate it will only return that if F_SETLKW is returned.
> 
> It should be harmless to use this wrapper if that's the case, though.

Yeah, I added it just in case the man page was incomplete.

[commit-bot: I haz the power, 2013-09-05 00:11:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dgrogan@chromium.org/23549017/1

[commit-bot: I haz the power, 2013-09-05 01:21:59]

Retried try job too often on ios_dbg_simulator for step(s) ui_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=ios_dbg_si...

[commit-bot: I haz the power, 2013-09-05 14:13:54]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dgrogan@chromium.org/23549017/1

[commit-bot: I haz the power, 2013-09-05 14:18:04]

Change committed as 221413