Refs #10260 OpenBSD support #25327

Refs #10260 OpenBSD support #25327

This branch adds support for the OpenBSD platform form issue #10260

For current status update please see:
https://github.com/dart-lang/sdk/issues/10260#issuecomment-168332554

At the time of opening this pull request:
 - the runtime can be built
 - the sdk can be built
 - 3780 passed; 5 failed; - see #10260 for details.
 - ongoing work to build the 'all' target

Full test run:
 - 13514 passed; 71 failed
 - SDK path failures are related to a still not implemented workaround for procfs

All tests & porting done on OpenBSD -current snapshot from Dec 19th 2015 amd64

---- changelog follows ----

Merge branch 'OpenBSD-support' into upstream-OpenBSD-support


Linter changes


Merge branch 'OpenBSD-support' into upstream-OpenBSD-support


More linter changes


Merge branch 'OpenBSD-support' into upstream-OpenBSD-support


Make the linter happy


Merge branch 'OpenBSD-support' into upstream-OpenBSD-support


Merge branch 'master' of https://github.com/dart-lang/sdk into OpenBSD-support


Include pthreads in the linking phase


Use Directory::SystemTemp from Linux instead of android hard coded path


Implement cpuinfo for OpenBSD using cpuid path from linux


Add OpenBSD to CheckedInSdkPath

This is used during the build for observatory_tool.py it will attempt
to download an SDK from google but there's none yet for OpenBSD.

This allows the build moving forward with a warning being printed that
this system can't run the SDK and that debug builds will be slow with
the bootstrap dart.

Assuming that upstream will pick up the port this should not be the case
in the future.

Enable building boringssl for OpenBSD using the linux sources


Add stdio_openbsd.cc based on stdio_android.cc


Implement a stub file system watcher for OpenBSD

There is no inotify like API available on OpenBSD

Follow upstream and add Utils::StrError based on Android implementation


Follow upstream extensions_openbsd changes


Follow upstream os_openbsd changes


Follow upstream by adding Platform::Exit for OpenBSD


Follow upstream refactorings

 - using Utils::StrError
 - renaming GetCurrentTimeMilliseconds -> GetCurrentMonotonicMillis etc.

Adjust to upstream os_thread & thread_interrupter changes


Merge branch 'master' into OpenBSD-support


Merge branch 'master' into OpenBSD-support


Add runtime/vm/native_symbol_openbsd.cc based on android


Add runtime/bin/utils_openbsd.cc based on android


Fix typo


Fix ifdef guard


Add runtime/vm/signal_handler_openbsd.cc based on linux

Use mcontext.sc_eip/rip etc instead of gregs

Add runtime/vm/virtual_memory_openbsd.cc based on android


Handle obtaining mcontext_t in thread_interrupter


Add runtime/vm/thread_interrupter_openbsd.cc based on linux


Add runtime/bin/fdutils_openbsd.cc based on fdutils_linux.cc


Implement debuginfo & gdbjit based on linux


Change if guardst


Add runtime/bin/process_openbsd.cc based on process_android.cc


Mark whole cpuinfo_openbsd.cc as UNIMPLEMENTED();


Add runtime/vm/cpuinfo_openbsd.cc based on cpuinfo_linux.cc


Add runtime/bin/extensions_openbsd.cc based on extensions_linux.cc


Add crypto_openbsd.cc and mark a FIXME for getentropy(2)


Include platform_openbsd.cc in build files


Handle environ for OpenBSD, add FIXME for procfs & minor tweaks

execve(2) makes environ available when a process begins
the handling matches what process_android does.

Add runtime/bin/platform_openbsd.cc based on platform_linux.cc


Proper OS::Name() & if preprocessor guard


Switch to posix_memalign & use getthrid() instead of a syscall


Add runtime/vm/os_openbsd.cc based on os_linux.cc


Reinterpret cast on ThreadIdFromIntPtr


Include os_thread_openbsd.{h,cc} in the build


Add runtime/bin/log_openbsd.cc based on log_linux.cc


Fix conditional include copy pasta


OpenBDS does have mkdtemp so drop the helper function


Add missing include for sys/socket.h


Add runtime/bin/file_openbsd.cc based on android


Add runtime/bin/directory_openbsd.cc based on android


More gypi sources to update


Add runtime/dart-runtime.gyp for openbsd based on linux


Drop -lrt & -ldl for OpenBSD in vm.gypi


Add OpenBSD to vm.gypi


Add runtime/bin/eventhandler_openbsd.{h,cc} based on macos for kqueue


Add runtime/bin/socket_openbsd.{h,cc} based on macos


Add runtime/bin/dbg_connection.{h,cc} based on macos for kqueue


Account for thread API differences between Android & OpenBSD


Add runtime/bin/thread_openbsd.cc based on thread_android.cc


Add runtime/bin/thread_openbsd.h based on thread_android.h


Merge branch 'master' into OpenBSD-support


Handle ucontext.h in signal_handler.h


Add runtime/vm/atomic.h & atomic_openbsd.h

Based on atomic_linux.h

Add platform/utils_openbsd.cc based on utils_linux.cc


Add os_thread for OpenBSD based on Android


Fix case for the OpenBSD platform


Define OpenBSD target OS in runtime/platform/globals.h


Make the build tool recognize OpenBSD

We use `gmake` like FreeBSD to use the GNU Make

Add configurations_openbsd.gypi

With this step we can generate Makefiles from a fresh gclient sync call:
 /usr/local/bin/python sdk/tools/gyp_dart.py

The file is a 1:1 copy of configurations_make.gypi which contained
Linux only targets but previous attempts at porting the software to
OpenBSD have shown that one can get really far with that setup using
those defaults.

Add OpenBSD as a dart target os

This opens up a can of worms of course but let's do things properly
and one small step at a time.

This change impacts:
 /usr/local/bin/python sdk/tools/gyp_dart.py

which is automatically called by gclient sync and now will faile with
missing configuration targets for Dart_OpenBSD_Base. Without the change
the error would just slap you with a generic unsupported os message.

One can of course set the target os to Linux which would get you much
further along with far less effort but that's a bit ugly solution.

BUG=

[ricow1, 2016-01-05 07:12:13]

ricow@google.com changed reviewers:
+ iposva@google.com, ricow@google.com

[ricow1, 2016-01-05 07:12:13]

+Ivan for actual review
Just a few fast comments:

It seems to me that a ton of the code can be shared between openbsd and mac
implementations, but I will defer to Ivan about that (I don't know how
interleaved he wants that code)

Replace FIXME with 
TODO(mulander): REASON

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
File runtime/bin/directory_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
runtime/bin/directory_openbsd.cc:346: // Android's getcwd adheres closely to the
POSIX standard. It won't
Android's?

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd.cc
File runtime/vm/cpuinfo_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:25: return;
why the return

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:31: return;
why the return?

https://codereview.chromium.org/1559053002/diff/1/tools/build.py
File tools/build.py (right):

https://codereview.chromium.org/1559053002/diff/1/tools/build.py#newcode111
tools/build.py:111: if not os_name in ['android', 'freebsd', 'linux', 'macos',
'openbsd', 'win32']:
long line

[ricow1, 2016-01-05 07:36:45]

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/bin.gypi#newcod...
runtime/bin/bin.gypi:905: }],
trailing whitespace

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi
File runtime/vm/vm.gypi (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode140
runtime/vm/vm.gypi:140: }],
trailing whitespace

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode210
runtime/vm/vm.gypi:210: }],
trailing whitespace

[mulander, 2016-01-05 11:31:03]

> +Ivan for actual review
> Just a few fast comments:

> It seems to me that a ton of the code can be shared between > openbsd and mac
> implementations, but I will defer to Ivan about that (I > don't know how
> interleaved he wants that code)

Regarding shared code that's not exactly true and requires some more
consideration.

OpenBSD reuses:
 - a lot of the android code as they share a libc
 - some Mac OS X code (mostly kqueue related)
 - a lot of linux code

FreeBSD reuses (from what I talked with the porter):
 - a lot of Mac OS X code
 - a lot of the Linux code

NetBSD (from what a developer told me):
 - reuses a lot of Linux
 - reuses a lot of MacOS X

While porting I did notice that the whole Android support is almost a verbatim
copy of Linux with a ton of code shared (and seemingly some recent changes not
applied to the Android files).

If there's plan to reuse that code I think it will require more common 'layers'
than just telling the build to share specific files.

Hope that's the right way to add a comment.

I'm starting to address code comments from this review - thanks for starting it!
:)

Regards,
Adam

[mulander, 2016-01-05 16:13:49]

Uploaded a new patch set addressing all raised issues. Sorry if it' a bit spammy
I'm still learning your code review tool :)

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/bin.gypi#newcod...
runtime/bin/bin.gypi:905: }],
On 2016/01/05 07:36:45, ricow1 wrote:
> trailing whitespace

Done.

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
File runtime/bin/directory_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
runtime/bin/directory_openbsd.cc:346: // Android's getcwd adheres closely to the
POSIX standard. It won't
On 2016/01/05 07:12:13, ricow1 wrote:
> Android's?

You are correct. getcwd(3) on OpenBSD does allocate the memory and the code will
be replaced with a single return getcwd(NULL, 0) call.

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
runtime/bin/directory_openbsd.cc:346: // Android's getcwd adheres closely to the
POSIX standard. It won't
On 2016/01/05 07:12:13, ricow1 wrote:
> Android's?

Acknowledged.

https://codereview.chromium.org/1559053002/diff/1/runtime/bin/directory_openb...
runtime/bin/directory_openbsd.cc:346: // Android's getcwd adheres closely to the
POSIX standard. It won't
On 2016/01/05 07:12:13, ricow1 wrote:
> Android's?

Done.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd.cc
File runtime/vm/cpuinfo_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:25: return;
On 2016/01/05 07:12:13, ricow1 wrote:
> why the return

Acknowledged. Not needed.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:25: return;
On 2016/01/05 07:12:13, ricow1 wrote:
> why the return

Done.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:31: return;
On 2016/01/05 07:12:13, ricow1 wrote:
> why the return?

Acknowledged.Not needed.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/cpuinfo_openbsd....
runtime/vm/cpuinfo_openbsd.cc:31: return;
On 2016/01/05 07:12:13, ricow1 wrote:
> why the return?

Done.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi
File runtime/vm/vm.gypi (right):

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode140
runtime/vm/vm.gypi:140: }],
On 2016/01/05 07:36:45, ricow1 wrote:
> trailing whitespace

Acknowledged.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode140
runtime/vm/vm.gypi:140: }],
On 2016/01/05 07:36:45, ricow1 wrote:
> trailing whitespace

Done.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode210
runtime/vm/vm.gypi:210: }],
On 2016/01/05 07:36:45, ricow1 wrote:
> trailing whitespace

Acknowledged.

https://codereview.chromium.org/1559053002/diff/1/runtime/vm/vm.gypi#newcode210
runtime/vm/vm.gypi:210: }],
On 2016/01/05 07:36:45, ricow1 wrote:
> trailing whitespace

Done.

https://codereview.chromium.org/1559053002/diff/1/tools/build.py
File tools/build.py (right):

https://codereview.chromium.org/1559053002/diff/1/tools/build.py#newcode111
tools/build.py:111: if not os_name in ['android', 'freebsd', 'linux', 'macos',
'openbsd', 'win32']:
On 2016/01/05 07:12:13, ricow1 wrote:
> long line

Acknowledged.

https://codereview.chromium.org/1559053002/diff/1/tools/build.py#newcode111
tools/build.py:111: if not os_name in ['android', 'freebsd', 'linux', 'macos',
'openbsd', 'win32']:
On 2016/01/05 07:12:13, ricow1 wrote:
> long line

Done.

[mulander, 2016-01-11 17:51:55]

It's been 6 days since any feedback. Is there anything else I need to do before
this change is merged? I would like to work on further changes for the platform
but without having this baseline change in I will be just bubbling up more
changes in this jumbo patch.

Regards,
Adam

[Ivan Posva, 2016-01-11 23:58:40]

I would like to understand why the files are name *_openbsd.[h|cc]. I do not
expect other BSD implementations to come with their own set of files. Instead I
would like for all of them to be unified.

Below is a first round of comments.

Thanks,
-Ivan

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/crypto_open...
File runtime/bin/crypto_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/crypto_open...
runtime/bin/crypto_openbsd.cc:20: // TODO(mulander): Consider using
getentropy(2)
Once you get there, please explain why you do use getentropy(2) and not
arc4random_buf(void *buf, size_t nbytes).

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
File runtime/bin/directory_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
runtime/bin/directory_openbsd.cc:367: 
Why? Generally we use two lines between functions.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
runtime/bin/directory_openbsd.cc:385: 
ditto

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
File runtime/bin/eventhandler_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
runtime/bin/eventhandler_openbsd.cc:9: #include "bin/eventhandler_openbsd.h"
This line is not needed as eventhandler.h will include eventhandler_openbsd.h. I
will fix the original files.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/extensions_...
File runtime/bin/extensions_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/extensions_...
runtime/bin/extensions_openbsd.cc:35: 
Please remove whitespace change.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/file_openbs...
File runtime/bin/file_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/file_openbs...
runtime/bin/file_openbsd.cc:291: int wrote = TEMP_FAILURE_RETRY(write(new_fd,
buffer, result));
Please add a TODO to deal with short writes. We will also have to deal with the
original code for Android and Linux. Filed http://dartbug.com/25399

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/platform_op...
File runtime/bin/platform_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/platform_op...
runtime/bin/platform_openbsd.cc:78: // TODO(mulander): There is no procfs on
OpenBSD - find a workaround
Would remembering arg[0] in main.cc work? It is not generally applicable for all
VM embedding scenarios, but could be a possibility here. Although I am not sure
whether this is even supported in OpenBSD.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/thread_open...
File runtime/bin/thread_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/thread_open...
runtime/bin/thread_openbsd.cc:9: #include "bin/thread_openbsd.h"
This line should not be necessary.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/utils_openb...
File runtime/bin/utils_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/utils_openb...
runtime/bin/utils_openbsd.cc:17: 
White space change?

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_openbsd.cc
File runtime/vm/os_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_openbsd.c...
runtime/vm/os_openbsd.cc:554: 
White space changes.

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
File runtime/vm/os_thread_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
runtime/vm/os_thread_openbsd.cc:110: 
White space changes.

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
File runtime/vm/os_thread_openbsd.h (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
runtime/vm/os_thread_openbsd.h:23: 
White space changes.

[mulander, 2016-01-12 00:22:46]

Regarding file naming, if you look at my previous comment -
https://codereview.chromium.org/1559053002/#msg4:

Regarding shared code that's not exactly true and requires some more
consideration.

OpenBSD reuses:
 - a lot of the android code as they share a libc
 - some Mac OS X code (mostly kqueue related)
 - a lot of linux code

FreeBSD reuses (from what I talked with the porter):
 - a lot of Mac OS X code
 - a lot of the Linux code

NetBSD (from what a developer told me):
 - reuses a lot of Linux
 - reuses a lot of MacOS X

---

I replied to code review commits, amending the code right now to address them.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/crypto_open...
File runtime/bin/crypto_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/crypto_open...
runtime/bin/crypto_openbsd.cc:20: // TODO(mulander): Consider using
getentropy(2)
On 2016/01/11 23:58:40, Ivan Posva wrote:
> Once you get there, please explain why you do use getentropy(2) and not
> arc4random_buf(void *buf, size_t nbytes).

You are correct. When that comment was made it was a quick note when I thought
about why I would need to open /dev/urandom if I have a syscall available. That
was some time ago... I did in fact intend to use arc4random_buf after seeing
this http://marc.info/?l=openbsd-tech&m=145176645017540&w=2 on the openbsd
mailing lists. So consider it an outdated comment. arc4random_buf is the target
implementation.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
File runtime/bin/directory_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
runtime/bin/directory_openbsd.cc:367: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> Why? Generally we use two lines between functions.

I was not aware of that. Will be fixed.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/directory_o...
runtime/bin/directory_openbsd.cc:385: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> ditto

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
File runtime/bin/eventhandler_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
runtime/bin/eventhandler_openbsd.cc:9: #include "bin/eventhandler_openbsd.h"
On 2016/01/11 23:58:40, Ivan Posva wrote:
> This line is not needed as eventhandler.h will include eventhandler_openbsd.h.
I
> will fix the original files.

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/extensions_...
File runtime/bin/extensions_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/extensions_...
runtime/bin/extensions_openbsd.cc:35: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> Please remove whitespace change.

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/file_openbs...
File runtime/bin/file_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/file_openbs...
runtime/bin/file_openbsd.cc:291: int wrote = TEMP_FAILURE_RETRY(write(new_fd,
buffer, result));
On 2016/01/11 23:58:40, Ivan Posva wrote:
> Please add a TODO to deal with short writes. We will also have to deal with
the
> original code for Android and Linux. Filed http://dartbug.com/25399

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/platform_op...
File runtime/bin/platform_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/platform_op...
runtime/bin/platform_openbsd.cc:78: // TODO(mulander): There is no procfs on
OpenBSD - find a workaround
On 2016/01/11 23:58:40, Ivan Posva wrote:
> Would remembering arg[0] in main.cc work? It is not generally applicable for
all
> VM embedding scenarios, but could be a possibility here. Although I am not
sure
> whether this is even supported in OpenBSD.

There is an example implementation in Go I want to have a closer look at:

https://github.com/kardianos/osext/commit/b4814f465fb1f92d46e37f7ef84d732ece7...

It's not the prettiest thing but it works pretty well for syncthing. arg[0] as
in what's passed to main should obviously work except any link / path magic
happening.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/thread_open...
File runtime/bin/thread_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/thread_open...
runtime/bin/thread_openbsd.cc:9: #include "bin/thread_openbsd.h"
On 2016/01/11 23:58:40, Ivan Posva wrote:
> This line should not be necessary.

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/utils_openb...
File runtime/bin/utils_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/utils_openb...
runtime/bin/utils_openbsd.cc:17: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> White space change?

My bad like in previous cases.

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_openbsd.cc
File runtime/vm/os_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_openbsd.c...
runtime/vm/os_openbsd.cc:554: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> White space changes.

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
File runtime/vm/os_thread_openbsd.cc (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
runtime/vm/os_thread_openbsd.cc:110: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> White space changes.

Acknowledged.

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
File runtime/vm/os_thread_openbsd.h (left):

https://codereview.chromium.org/1559053002/diff/40001/runtime/vm/os_thread_op...
runtime/vm/os_thread_openbsd.h:23: 
On 2016/01/11 23:58:40, Ivan Posva wrote:
> White space changes.

Acknowledged.

[mulander, 2016-01-12 00:59:39]

This issue will not be addressed as git cl upload prevents me from doing it.

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
File runtime/bin/eventhandler_openbsd.cc (right):

https://codereview.chromium.org/1559053002/diff/40001/runtime/bin/eventhandle...
runtime/bin/eventhandler_openbsd.cc:9: #include "bin/eventhandler_openbsd.h"
On 2016/01/11 23:58:40, Ivan Posva wrote:
> This line is not needed as eventhandler.h will include eventhandler_openbsd.h.
I
> will fix the original files.

I can't address this issue as git cl upload prevents me from uploading:
/home/mulander/github.com/mulander/sdk/runtime/bin/eventhandler_openbsd.cc:5: 
runtime/bin/eventhandler_openbsd.cc should include its header file
runtime/bin/eventhandler_openbsd.h  [bui$
d/include] [5]
/home/mulander/github.com/mulander/sdk/runtime/bin/thread_openbsd.cc:5: 
runtime/bin/thread_openbsd.cc should include its header file
runtime/bin/thread_openbsd.h  [build/include] [5]