Initial WebAssembly support.

[waywardmonkeys]

Fixes #168.

[waywardmonkeys]

I meant for this to be a draft. I will be pushing or force-pushing to it over the coming days / weeks.

[waywardmonkeys]

This is a more interesting build error, haven't looked into it yet, requires changes I haven't pushed yet to get to this point:

eventtxt.c:162:3: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
  EVENT_CLOCK_MAKE(val, low, high);
  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
./clock.h:162:35: note: expanded from macro 'EVENT_CLOCK_MAKE'
  ((lvalue) = ((EventClock)(high) << 32) + ((EventClock)(low) & (0xfffffffful)))
                                  ^  ~~
1 error generated.

[rptb1]

This is a more interesting build error

Are you not getting a 64-bit unsigned long long?

mps/code/clock.h

Line 111 in 142999b

__extension__ typedef unsigned long long EventClock;

[waywardmonkeys]

That's line 111, the error was on line 162, which is the mps_clock_t code path ... mps_clock_t is mps_word_t

[rptb1]

Ah yes. Looks like a 64-bit assumption has crept in here

mps/code/clock.h

Lines 156 to 162 in 142999b

	/* no fast clock, use plinth, probably from the C library */
	#ifndef EVENT_CLOCK
	typedef mps_clock_t EventClock;
	#define EVENT_CLOCK_MAKE(lvalue, low, high) \
	((lvalue) = ((EventClock)(high) << 32) + ((EventClock)(low) & (0xfffffffful)))

but since

mps/code/mps.h

Line 77 in 142999b

typedef mps_word_t mps_clock_t; /* processor time */

we could switch here on MPS_WORD_WIDTH or define MPS_CLOCK_WIDTH to be MPS_WORD_WIDTH and switch on that.

[rptb1]

EVENT_CLOCK_MAKE is only used in

mps/code/eventtxt.c

Lines 152 to 166 in 142999b

	static EventClock parseClock(char **pInOut)
	{
	EventClock val;
	int i, l;
	unsigned long low, high;
	char *p = *pInOut;
	i = sscanf(p, "%08lX%08lX%n", &high, &low, &l);
	if (i != 2)
	everror("Couldn't read a clock from '%s'", p);
	EVENT_CLOCK_MAKE(val, low, high);
	*pInOut = p + l;
	return val;
	}

which is intended to only be run on the same platform as the MPS that generated the telemetry, though we might be able to make it more flexible.

[thejayps]

JPH+RB: This needs CI before merging but currently not a high priority to merge, not currently impacting customers. Assigning importance "optional".

[waywardmonkeys]

JPH+RB: This needs CI before merging but currently not a high priority to merge

FWIW, it does have some CI... but needs more. But there was an issue with the CI for it because it was building the threaded tests as well ...