Switch the nop pnacl x86-64 IRT shim to the real one generated from IDL.

pnacl/build.sh

Index: pnacl/build.sh
diff --git a/pnacl/build.sh b/pnacl/build.sh
index 64f2a07a4271ee937f47fa9764c0406086367304..3ca8c4e6a5f3effbf5ceb59ecc8e57179eadc210 100755
--- a/pnacl/build.sh
+++ b/pnacl/build.sh
@@ -3048,13 +3048,17 @@ sdk-irt-shim() {
   #       the pnacl toolchain to depend on it.
   StepBanner "SDK" "IRT Shim"
   spushd "${NACL_ROOT}"
+  # NOTE: We specify bitcode=1, but it is really using nacl-gcc to build
+  # the library (it's only bitcode=1 because it's part of the pnacl sdk).
   RunWithLog "sdk.irt.shim" \

[robertm, 2011/12/01 22:55:43]

this looks like churn why not stick with the original

[jvoung - send to chromium..., 2011/12/01 23:35:34]

I think this was to make the implicit-libs change work, which makes it so that
when you run:

./scons bitcode=1 platform=x86-64 chrome_browser_tests

It would rebuild the latest libpnacl_irt_shim.a in the pnacl-clang directory
before linking together the browser nexes and use that, rather than the
libpnacl_irt_shim.a that was downloaded from the tarball (built here).

We could leave it like the original, but then we'd need to do a toolchain DEPS
update before it gets the newer shim.

Thoughts?

[robertm, 2011/12/01 23:58:43]

Ok that is fine then - but  maybe add a brief comment

On 2011/12/01 23:35:34, jvoung wrote:

     ./scons -j${PNACL_CONCURRENCY} \
+            bitcode=1 \
             platform=x86-64 \
             naclsdk_validate=0 \
             --verbose \
             pnacl_irt_shim
-  local outdir="${SCONS_OUT}"/nacl-x86-64/obj/src/untrusted/pnacl_irt_shim
+  local out_dir_prefix="${SCONS_OUT}"/nacl-x86-64-pnacl-clang
+  local outdir="${out_dir_prefix}"/obj/src/untrusted/pnacl_irt_shim
   mkdir -p "${INSTALL_LIB_X8664}"
   cp "${outdir}"/libpnacl_irt_shim.a "${INSTALL_LIB_X8664}"
   spopd