getPressure: handle error when we don't have an exception

[bobrippling]

This handles an I2C error, originally ocurring in #2137, found again by chance in an unrelated PR.

When we create a promise for getPressure, we now look at both jspGetException and then jspHasError (separately).
If the former is null but we have an error, we raise an exception about the I2C communication with the barometer.

Previous behaviour would raise an exception with no payload, leading to the Uncaught Error: Unhandled promise rejection: undefined seen in the BangleApps pull request above.

[bobrippling]

I believe I have the locking correct here - jswrap_internalerror_constructor unlocks msg, but would appreciate a second pair of eyes

[gfwilliams]

I don't think jswrap_internalerror_constructor does unlock msg - if it's a function called from JS (which I'm pretty sure it is as it's the constructor) then the caller is responsible for unlocking

[bobrippling]

I might've misread the code - I think here it unlocks in jsvObjectSetChildAndUnLock

Espruino/src/jswrap_error.c

Lines 51 to 62 in deedeaf

	JsVar *_jswrap_error_constructor(JsVar *msg, char *type) {
	JsVar *d = jspNewObject(0,type);
	if (!d) return 0;
	if (msg) {
	msg = jsvAsString(msg);
	jsvObjectSetChildAndUnLock(d, "message", msg);
	}
	jsvObjectSetChildAndUnLock(d, "type", jsvNewFromString(type));
	return d;
	}

(⬆️ msg here)

Espruino/src/jsvar.c

Lines 3195 to 3198 in deedeaf

	void jsvObjectSetChildAndUnLock(JsVar *obj, const char *name, JsVar *child) {
	jsvObjectSetChild(obj, name, child);
	jsvUnLock(child);
	}

(⬆️ child here)

[gfwilliams]

Ahh - jsvObjectSetChildAndUnLock(d, "message", msg); does unlock msg but on the line above, msg = jsvAsString(msg); replaces msg with a newly locked version of msg that's guaranteed to be a string. I guess for readability it should really be written to a new variable instead.

[bobrippling]

That makes sense - I could make that change as part of this PR if you like?

[gfwilliams]

No, it's ok - i'll do it separately now...

[gfwilliams]

but you still need to unlock msg...

[bobrippling]

Ah yes - I was assuming jsvAsString unlocked the value it was given, with us replacing it. Thanks for the pointer!

[gfwilliams]

Looks like there's a build error here as well? https://github.com/espruino/Espruino/actions/runs/7112131428/job/19361472245?pr=2438

edit: ahh - you made a comment about this above

[bobrippling]

This is ready for another look btw :)

[gfwilliams]

Thanks - just looked again and this looks good! Merging now!