PNaCl: Allow translator to produce an unsandboxed, native executable

PNaCl: Allow translator to produce an unsandboxed, native executable

This only works for targeting x86-32 Linux in this initial version.
This allows pnacl-translate to be invoked with "-arch linux-x86-32".
I'll extend this to work on x86-64 and Mac in later changes.

 * Add a minimal IRT implementation that's just enough to run
   hello_world.pexe.  It implements TLS, mmap(), write(), exit(), etc.
   I'll implement more IRT interfaces in a later change.

 * For testing, add a Buildbot step to test this, and add the
   "pnacl_unsandboxed" option to Scons.

BUG=https://code.google.com/p/nativeclient/issues/detail?id=3699
TEST=run_hello_world_test_irt bitcode=1 pnacl_unsandboxed=1 --mode=nacl_irt_test

Committed: https://src.chromium.org/viewvc/native_client?view=rev&revision=12218

[Derek Schuff, 2013-10-07 20:30:30]

On 2013/10/07 15:43:50, Mark Seaborn wrote:

It seems weird to pack the OS/ABI into pnacl-translate's arch flag. Maybe we
should just make the translation target a triple (or just have a separate
variable for OS) and give pnacl-translate an OS flag which defaults to nacl.

[Derek Schuff, 2013-10-07 20:30:35]

https://codereview.chromium.org/25499003/diff/6001/buildbot/buildbot_pnacl.sh
File buildbot/buildbot_pnacl.sh (right):

https://codereview.chromium.org/25499003/diff/6001/buildbot/buildbot_pnacl.sh...
buildbot/buildbot_pnacl.sh:611: ./scons run_hello_world_test_irt bitcode=1
pnacl_unsandboxed=1 \
would it make more sense to run this on the FYI bot? we aren't depending on this
for any real use yet, are we?

[Derek Schuff, 2013-10-07 21:03:31]

lgtm

[eliben_google.com, 2013-10-07 21:08:03]

On Mon, Oct 7, 2013 at 8:43 AM, <mseaborn@chromium.org> wrote:

> Reviewers: Derek Schuff,
>
> Description:
> PNaCl: Allow translator to produce an unsandboxed, native executable
>
> This only works for targeting x86-32 Linux in this initial version.
> This allows pnacl-translate to be invoked with "-arch linux-x86-32".
> I'll extend this to work on x86-64 and Mac in later changes.
>
>  * Add a minimal IRT implementation that's just enough to run
>    hello_world.pexe.  It implements TLS, mmap(), write(), exit(), etc.
>    I'll implement more IRT interfaces in a later change.
>
>  * For testing, add a Buildbot step to test this, and add the
>    "pnacl_unsandboxed" option to Scons.
>

Is this starting the implementation for task "BH    Translator producing
sandbox-less executables" from the brainstorming session, or unrelated?

-- 
You received this message because you are subscribed to the Google Groups
"Native-Client-Reviews" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to native-client-reviews+unsubscribe@googlegroups.com.
To post to this group, send email to native-client-reviews@googlegroups.com.
Visit this group at http://groups.google.com/group/native-client-reviews.
For more options, visit https://groups.google.com/groups/opt_out.

[Mark Seaborn, 2013-10-07 21:08:53]

On 7 October 2013 14:07, Eli Bendersky <eliben@google.com> wrote:

> On Mon, Oct 7, 2013 at 8:43 AM, <mseaborn@chromium.org> wrote:
>
>> Reviewers: Derek Schuff,
>>
>> Description:
>> PNaCl: Allow translator to produce an unsandboxed, native executable
>>
>> This only works for targeting x86-32 Linux in this initial version.
>> This allows pnacl-translate to be invoked with "-arch linux-x86-32".
>> I'll extend this to work on x86-64 and Mac in later changes.
>>
>>  * Add a minimal IRT implementation that's just enough to run
>>    hello_world.pexe.  It implements TLS, mmap(), write(), exit(), etc.
>>    I'll implement more IRT interfaces in a later change.
>>
>>  * For testing, add a Buildbot step to test this, and add the
>>    "pnacl_unsandboxed" option to Scons.
>>
>
> Is this starting the implementation for task "BH    Translator producing
> sandbox-less executables" from the brainstorming session
>

Yes, it is.

Mark

-- 
You received this message because you are subscribed to the Google Groups
"Native-Client-Reviews" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to native-client-reviews+unsubscribe@googlegroups.com.
To post to this group, send email to native-client-reviews@googlegroups.com.
Visit this group at http://groups.google.com/group/native-client-reviews.
For more options, visit https://groups.google.com/groups/opt_out.

[Mark Seaborn, 2013-10-08 18:23:52]

On 7 October 2013 13:30, <dschuff@chromium.org> wrote:

> https://codereview.chromium.**org/25499003/diff/6001/**
>
buildbot/buildbot_pnacl.sh<https://codereview.chromium.org/25499003/diff/6001/buildbot/buildbot_pnacl.sh>
> File buildbot/buildbot_pnacl.sh (right):
>
> https://codereview.chromium.**org/25499003/diff/6001/**
>
buildbot/buildbot_pnacl.sh#**newcode611<https://codereview.chromium.org/25499003/diff/6001/buildbot/buildbot_pnacl.sh#newcode611>
> buildbot/buildbot_pnacl.sh:**611: ./scons run_hello_world_test_irt
> bitcode=1 pnacl_unsandboxed=1 \
> would it make more sense to run this on the FYI bot? we aren't depending
> on this for any real use yet, are we?


My point of view is that the FYI bots are for tests that are too slow to
run on the toolchain trybots, not for tests that we mind breaking.  Even if
a feature is experimental, ideally its tests should run on the trybots --
it's just frustrating to have code in the codebase that gets broken
accidentally for lack of tests.

It seems weird to pack the OS/ABI into pnacl-translate's arch flag. Maybe we
> should just make the translation target a triple (or just have a separate
> variable for OS) and give pnacl-translate an OS flag which defaults to
> nacl.


I think potentially it's convenient to be able to pass a single string in
"-arch", as with target triples.  This could become more triple-like in
future.

For internal use inside pnacl-translate.py, I agree that separate fields
would be cleaner, to reduce code duplication.  I can refactor this when
extending this to Mac, x86-64, etc.

Cheers,
Mark

-- 
You received this message because you are subscribed to the Google Groups
"Native-Client-Reviews" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to native-client-reviews+unsubscribe@googlegroups.com.
To post to this group, send email to native-client-reviews@googlegroups.com.
Visit this group at http://groups.google.com/group/native-client-reviews.
For more options, visit https://groups.google.com/groups/opt_out.

[Mark Seaborn, 2013-10-08 18:27:26]

Committed patchset #2 manually as r12218 (presubmit successful).