Add support for generating nmfs for pnacl pexes with glibc (vs pnacl nexes).

Add support for generating nmfs for pnacl pexes with glibc (vs pnacl nexes).
This currently hooks onto the scons stuff where a base nmf file is required.

BUG= http://code.google.com/p/nativeclient/issues/detail?id=2546
TEST= run_pnacl_example_browser_test with --nacl_glibc, and it gets past the nmf generation stage.

[jvoung - send to chromium..., 2012-01-20 18:47:10]

ptal.

Sample output from scons staging directory with platform=x86-32 bitcode=1
--nacl_glibc:


{
"program":{
  "x86-64":{ "url":"runnable-ld.so"},
  "arm":{ "url":"runnable-ld.so"},
  "x86-32":{ "url":"runnable-ld.so"}
  },
"files":{
  "main.nexe":{
    "portable":{
    "url":"pnacl_test_example.pexe"}
    },
  "libm.so.3c8d1f2e":{
    "x86-32":{
    "url":"libm.so.3c8d1f2e"}
    },
  "libstdc++.so.6":{
    "x86-32":{
    "url":"libstdc++.so.6"}
    },
  "ld-nacl-x86-32.so.1":{
    "x86-32":{
    "url":"ld-nacl-x86-32.so.1"}
    },
  "libc.so.3c8d1f2e":{
    "x86-32":{
    "url":"libc.so.3c8d1f2e"}
    },
  "libgio.so":{
    "x86-32":{
    "url":"libgio.so"}
    },
  "libplatform.so":{
    "x86-32":{
    "url":"libplatform.so"}
    },
  "libpthread.so.3c8d1f2e":{
    "x86-32":{
    "url":"libpthread.so.3c8d1f2e"}
    }
  }
}

http://codereview.chromium.org/9264026/diff/13002/tests/pnacl_client_translat...
File tests/pnacl_client_translator/pnacl_test_example.html (right):

http://codereview.chromium.org/9264026/diff/13002/tests/pnacl_client_translat...
tests/pnacl_client_translator/pnacl_test_example.html:55: <embed
type="application/x-nacl" id="test_nexe2"
Well, now we are launching 6 nexes simultaneously =)

But we get some slight coverage on a different nmf format w.r.t the nmf
"generator" in scons...

[pdox, 2012-01-20 21:48:57]

I would strongly prefer we either:

1) Emit a list of strings from pnacl-meta, and parse that list in SConstruct,
instead of emulating the ldd output format from pnacl-meta.

-or-

2) Add a new script, pnacl-nmf.py, for building nmf files from a pexe. Both
SCons and PNaCl users could use this.

But I'll defer to other reviewers.

[jvoung - send to chromium..., 2012-01-23 22:31:00]

Went with 2) -- It theoretically traces .pso's for more dependencies until no
more are found, and does not attempt to trace .so's (we could extend it to run
objdump and grep the NEEDED lines of output or something though)...



On 2012/01/20 21:48:57, pdox wrote:
> I would strongly prefer we either:
> 
> 1) Emit a list of strings from pnacl-meta, and parse that list in SConstruct,
> instead of emulating the ldd output format from pnacl-meta.
> 
> -or-
> 
> 2) Add a new script, pnacl-nmf.py, for building nmf files from a pexe. Both
> SCons and PNaCl users could use this.
> 
> But I'll defer to other reviewers.

[jvoung - send to chromium..., 2012-01-24 00:03:21]

Hmm can't really make a .pso test case right now, so I think this is about it
for now. PTAL.

re: making a .pso test case, see --
http://code.google.com/p/nativeclient/issues/detail?id=2556



On 2012/01/23 22:31:00, jvoung wrote:
> Went with 2) -- It theoretically traces .pso's for more dependencies until no
> more are found, and does not attempt to trace .so's (we could extend it to run
> objdump and grep the NEEDED lines of output or something though)...
> 
> 
> 
> On 2012/01/20 21:48:57, pdox wrote:
> > I would strongly prefer we either:
> > 
> > 1) Emit a list of strings from pnacl-meta, and parse that list in
SConstruct,
> > instead of emulating the ldd output format from pnacl-meta.
> > 
> > -or-
> > 
> > 2) Add a new script, pnacl-nmf.py, for building nmf files from a pexe. Both
> > SCons and PNaCl users could use this.
> > 
> > But I'll defer to other reviewers.

[sehr (please use chromium), 2012-01-26 16:38:36]

On 2012/01/24 00:03:21, jvoung wrote:
> Hmm can't really make a .pso test case right now, so I think this is about it
> for now. PTAL.
> 
> re: making a .pso test case, see --
> http://code.google.com/p/nativeclient/issues/detail?id=2556
> 
> 
> 
> On 2012/01/23 22:31:00, jvoung wrote:
> > Went with 2) -- It theoretically traces .pso's for more dependencies until
no
> > more are found, and does not attempt to trace .so's (we could extend it to
run
> > objdump and grep the NEEDED lines of output or something though)...
> > 
> > 
> > 
> > On 2012/01/20 21:48:57, pdox wrote:
> > > I would strongly prefer we either:
> > > 
> > > 1) Emit a list of strings from pnacl-meta, and parse that list in
> SConstruct,
> > > instead of emulating the ldd output format from pnacl-meta.
> > > 
> > > -or-
> > > 
> > > 2) Add a new script, pnacl-nmf.py, for building nmf files from a pexe.
Both
> > > SCons and PNaCl users could use this.
> > > 
> > > But I'll defer to other reviewers.

LGTM.

[jvoung - send to chromium..., 2012-01-26 18:24:34]

On 2012/01/26 16:38:36, sehr wrote:
> On 2012/01/24 00:03:21, jvoung wrote:
> > Hmm can't really make a .pso test case right now, so I think this is about
it
> > for now. PTAL.
> > 
> > re: making a .pso test case, see --
> > http://code.google.com/p/nativeclient/issues/detail?id=2556
> > 
> > 
> > 
> > On 2012/01/23 22:31:00, jvoung wrote:
> > > Went with 2) -- It theoretically traces .pso's for more dependencies until
> no
> > > more are found, and does not attempt to trace .so's (we could extend it to
> run
> > > objdump and grep the NEEDED lines of output or something though)...
> > > 
> > > 
> > > 
> > > On 2012/01/20 21:48:57, pdox wrote:
> > > > I would strongly prefer we either:
> > > > 
> > > > 1) Emit a list of strings from pnacl-meta, and parse that list in
> > SConstruct,
> > > > instead of emulating the ldd output format from pnacl-meta.
> > > > 
> > > > -or-
> > > > 
> > > > 2) Add a new script, pnacl-nmf.py, for building nmf files from a pexe.
> Both
> > > > SCons and PNaCl users could use this.
> > > > 
> > > > But I'll defer to other reviewers.
> 
> LGTM.

Ah my git branch somehow got corrupted or something:

fatal: failed to read object ...

Will re-upload.