targets/zephyr: Update to new Jerry API and improve REPL interaction. #1190
Conversation
|
This follows Unix port (main-unix.c) and API example doc. |
|
@sergioamr, @poussa: FYI. |
|
Sounds like a good idea. The run_simple initializes and shuts down the engine, so it cannot return with any meaningful value, since jerry heap has been freed already. |
|
LGTM |
1 similar comment
|
LGTM |
|
@LaszloLango : If #1177 is close to be merged (it seems to be), I guess it makes sense to land it first, then I'll update this patch. Thanks. |
pfalcon
changed the title
|
LGTM. I would like to be able to have the Travis integration working, so changes on the API get highlighted and do not break the build. The situation is that @Akromx16 will go back to university really soon and i am not sure if he would be able to finish his integration on #1144 Could someone with travis experience help on that PR? |
|
#1177 is merged. |
LaszloLango
added
the
jerry-port
pfalcon
changed the title
|
@LaszloLango : Thanks, refactored to the new API, rebased to master as of now. |
|
LGTM |
| @@ -158,6 +158,7 @@ const struct shell_cmd commands[] = | |||
| void main (void) | |||
| { | |||
| printf ("Jerry Compilation " __DATE__ " " __TIME__ "\n"); | |||
| jerry_init (JERRY_INIT_EMPTY); | |||
| shell_register_app_cmd_handler (shell_cmd_handler); | |||
| shell_init ("js> ", commands); | |||
| } /* main */ | |||
There was a problem hiding this comment.
jerry_cleanup should be called before the end of main
There was a problem hiding this comment.
@LaszloLango : That's not the case with Zephyr. Generally, a typical embedded app would work as an infinite loop (it "exits" via hardware reset), and specifically with Zephyr, shell_init() just sets up background task and exit, all actual input passed to callbacks at later time. So, it's not ok to deinitialize Jerry after call to shell_init(), and per above, no need to deinitialize at all. Hope that makes sense. I'm adding a comment to the patch with this info.
There was a problem hiding this comment.
shell_init will spawn a fiber that performs the shell task.
task_fiber_start(stack, STACKSIZE, shell, 0, 0, 7, 0);
If you clean zephyr there it will just crash.
With the change of running code on eval i think pfalcon is right on not cleaning up after execution.
There was a problem hiding this comment.
@pfalcon, @sergioamr, OK, thanks for the explanation.
|
LGTM after |
|
lgtm |
Various calls updated to use API from jerry-api.h (to-be JerryScript 1.0
API).
Use jerry_eval() instead of jerry_run_simple(), as it allow to store (and
then access) variables in a global environment. E.g. following now works
(entered and executed as two separate lines):
a = 1
print(a)
JerryScript-DCO-1.0-Signed-off-by: Paul Sokolovsky [email protected]
|
Added comment to the source why jerry_cleanup() is not called, to address @LaszloLango's comment above. |
|
LGTM |
Using jerry_eval() instead of jerry_run_simple() allows to store (and then access)
variables in a global environment. E.g. following now works:
JerryScript-DCO-1.0-Signed-off-by: Paul Sokolovsky [email protected]