Stack arguments wrong for <JS Function InstantiateFunction> - trace output differs on PPC

[andrewlow]

Found this while trying to resolve NPM issues.

Build node on Intel & PPC the same way

PPC

./configure --debug --dest-cpu=ppc; make -j20

Intel

./configure --debug --dest-cpu=ia32; make -j7

Now run the following on both.

./node --trace deps/npm/bin/npm-cli.js install url | head -2

Intel output

$ ./node --trace deps/npm/bin/npm-cli.js install url | head -2 1: Instantiate+59(this=0x31f29af9 <JS Object>, 0x31f34711 <FunctionTemplateInfo>, 0x31f08091 <undefined>) { 2: InstantiateFunction+64(this=0x31f29af9 <JS Object>, 0x31f34711 <FunctionTemplateInfo>, 0x31f08091 <undefined>) {

PPC output

$ ./node --trace deps/npm/bin/npm-cli.js install url | head -2 1: Instantiate+88(this=0x9a829ae9 <JS Object>, 0x9a8346e5 <FunctionTemplateInfo>, 0x9a808091 <undefined>) { 2: InstantiateFunction+108(this=0x9a808091 <undefined>, 0x9a8346e5 <FunctionTemplateInfo>, 0x9a808091 <undefined>) {

If you look carefully - you'll notice that the value of this should be <JS Object> but is <undefined> on PPC. I'm sort of surprised anything works.

[andrewlow]

It appears that the ARM simulator also exhibits the same issue. Since the PPC V8 port started from the ARM as a base, it appears we've duplicated this oddity.

I wonder if this has any impact on the accuracy of the runtime? I may be chasing a red-herring.

[andrewlow]

I'm going to close this - clearly it is not causing problems with any of the tests.