[NaCl SDK] Remove create_nmf dependency on NACL_SDK_ROOT env var

third_party/liblouis/liblouis_nacl.gyp

Index: third_party/liblouis/liblouis_nacl.gyp
diff --git a/third_party/liblouis/liblouis_nacl.gyp b/third_party/liblouis/liblouis_nacl.gyp
index 82336b86ef0cd8be5fdca74e9f91815b1da5097d..081658fba3c2f31583d35b6af487d5d985a21d35 100644
--- a/third_party/liblouis/liblouis_nacl.gyp
+++ b/third_party/liblouis/liblouis_nacl.gyp
@@ -138,6 +138,8 @@
                 '<(DEPTH)/native_client_sdk/src/tools/create_nmf.py',
                 '>@(_inputs)',
                 '--output=>(nmf)',
+                '--no-default-libpath',

[binji, 2015/04/07 19:44:02]

it seems that this change is the only difference between this CL and the
previous reverted CL. Do the trybots test this path? It seems that they
succeeded before without this change, so I would assume not.

[Sam Clegg, 2015/04/07 20:11:53]

No, I guess the bots don't test this otherwise the previous change would not
have landed.

I tested this by building this target locally.

+                '--objdump=>(nacl_glibc_tc_root)/bin/x86_64-nacl-objdump',

[binji, 2015/04/07 21:17:39]

does this need to be nacl_newlib_tc_root? Shouldn't make a difference I'd think,
but this build is newlib not glibc. Perhaps the toolchain won't be prepped if it
is built in isolation?

[Sam Clegg, 2015/04/09 00:26:01]

It won't actually use the objdump executable unless the nexe is dynamically
linked, so I think we are good.

               ],
             },
           ],