My Sky

A secret garden of programming and random things

[picsthatmakeyougohmm]

[a1291762]

[existencebringsonlypain]

i need data for a statistics project for school, so be my sample data, worms. i need thirty people minimum so if there aren't enough voters yet i'd love if you could help. thank you very much. worms.

take this test (https://www.keithcirkel.co.uk/whats-my-jnd/), then come back here:

what's your JND?

.00030-.00099

.0010-.0017

.0017-.0024

.0024-.0031

.0031-.0038

.0038-.0045

.0045-.0052

.0052-.0059

.0059-.0066

.0066-.0073

.0073-.0080

.0080 or greater

Voting ended onMay 13

it doesnt have to be a good score, you dont have to take it multiple times, you dont have to get on a good screen, etcetera. just gimme your score please this is my final project grade :)

i'd love if you could reblog for reach

[a1291762]

I thought my colour blindness would sink me but apparently not. I have very poor sensitivity to red.

Some screens I had no idea but most screens I could see the line.

One screen I clicked randomly and got it right anyway.

0.0084

[a1291762]

The Division Resurgence was supposed to be a mobile game but they've gone and released it for PC.

Not that I'm complaining. The touchscreen controls were janky and I'd already found ways to run from the phone mirrored to the PC monitor as well as run in an emulator.

Naturally the native PC version runs so much better than either of those options. It's the same mobile-focussed game. Just with better graphics.

[why-the-heck-not]

why are long escalators so scary. 172829 different scenarios going on in my brain about how i’m going to trip and domino-effect kill 263 ppl and injure 745 more on my way down

[a1291762]

Longer escalators are scary because the failure mode is so catastrophic.

2018 Rome escalator accident - Wikipedia

If you want alarming visuals.

[a1291762]

The Division Resurgence is clearly optimised to run on low end phones. That's cool, but it shows sometimes. Like the draw distance. It's actually quite close. Most of the time you won't notice it because there's stuff in the way but with some elevation it becomes clear. Here's me standing on an ambulance.

You'll need to zoom that in to see it. Here, use my scope.

That is not road. It's weird lack of texture. It's even worse when I turn right.

It's a veritable sea of un-textured road out there.

[a1291762]

In the division resurgence, this unassuming wall on the side of the main base should not be climbable, yet it is.

Once up there you can go right, into an area you're clearly not supposed to access. This is I think technically below the base.

[a1291762]

If you can tolerate bolero, this chiptune version is great. If nearly 15 minutes of the same riff repeated over and over, with progressive instrumentation isn't for you, don't bother.

[machine-saint]

one of the small joys of writing code is when you need an arbitrary string for a test case and you get to pick your favorite pop culture reference or pet's name or whatever

[a1291762]

Work product has networking. All the test bridge interfaces have the names of actual bridges given to them.

Similar jokes abound in the test code. Like the bond interface called 007.

[a1291762]

I had to make some choices, when translating the full colour source to the black and white screen but I'm happy with how it came out.

[charyou-tree]

When you're debugging code, there are a number of possible scenarios, in order of increasing scariness:

The code works, but with some known issues.

The code doesn't work, but I know why.

The code doesn't work, and I don't know why.

The code used to work, but I know what changed.

The code used to work, and I don't know what changed.

The code used to work, failed, and I fixed it so it works again, but now that I fixed it I'm not sure how it EVER worked before.

The code works, even though it clearly shouldn't, and I don't know why.

Just fixed a #6 issue at work so I'm a little shook.

[foone]

Don't forget

8. The code NEVER worked, but some other code was also broken, making you incorrectly think the code worked.

[a1291762]

I noticed a problem. The drill mast angles were being measured back to front. But when I fixed it, things stopped working.

Eventually I found another place that was also incorrectly flipping the angle direction.

Sometimes, two wrongs do make a right!

[a1291762]

The old config stack has some choices. One of these is that allocated objects are tied to a session. But what if you need to validate the entire database? The answer is it ends up allocating a huge amount of RAM! This is not great so I set out to fix it.

I used valgrind massif to find the worst offences and ended up with some nifty logic to throw things away early. The system does a depth first search during validation so as you come back up you can drop whole trees.

There were a few curve balls but nothing the tests, valgrind and logs didn't reveal. Updates are special because validation runs and then the database is updated so all the existing elements need to stay rather than being freed during validation.

However, a migration that just went in was causing the system to blow up. The logic wasn't complicated, delete a field that was set wrong, allowing the default value to be used instead. But when the database was updated, the deleted value was written instead of the default value.

Lots of debugging later and I tracked it down. When you delete an element it remains but is marked. When the validation logic runs the field is noted as required but absent and a new, anonymous element is created with the default value. Because it didn't exist before validation, this temporary element was being freed by my new logic. But this is an update operation so this element needed to be kept around!

After validation there is some ... resolution code? The code literally references Thunderdome. 2 elements come in, one emerges the winner. In this case, the deleted element with the bad value and the element with the default value. How these 2 got linked to each other is a mystery, but now that the default value isn't being freed early, it's working correctly.

[a1291762]

Time, date, seconds, battery, health (steps, minutes of exercise, active hours). It's coming together.

I originally wrote this watch face for a TicWatch E, which ran Android Wear and had a 400x400 (round) screen.

I ported it to the TicWatch E3, which had a 360x360 (round) screen. Since all my layout code was in pixels, I added a px function and wrapped all my constants. Then I calculated an appropriate scale factor and used it in the px function. Gross but it worked.

That same px function is still here, so this UI is still designed around a 400x400 display. Except that I have made a bunch of tweaks, threw some numbers into the corners. And the bounds actually go off the screen. I probably need to retire the scaling and just re-do the layout for this much smaller display.

But I do want this to work on the larger Pebble 2 Time when it arrives so some kind of scale factor may still be required...

[orteil42]

Wait, you were actually born in the 1900's? Thats so cool

i am going to eat my own entire skin

[idabbleincrazy]

Reblog if you were born in the 1900's.

[avoiding-claws]

[alexaloraetheris]

[kittydesade]

[gothiccharmschool]

[froschimbrunnen]

[ms-nerd]

[wizzardclown]

[virtual-particle]

[a1291762]

In a pebble app, you can include math.h and invoke sin, but it crashes with some values. Turns out you're supposed to use their special sin function, which is cool and all, but now my ported-from-Android drawing code needs a helper function. Ditto for cos.

Pebble rects use the sane x y w h rather than Androids insane x1 y1 x2 y2.

Drawing an arc is a little different to Android but a one-line helper function makes it the same. But strokes have rounded ends, not straight ends.

[a1291762]

I loaded tumblr on my computer so I could post some code...

Here's my sin to pebble sin adapter function.

float math_sin(float radians) { float degrees = radians * 180 / PI; int32_t _angle = DEG_TO_TRIGANGLE(degrees); int32_t _ret = sin_lookup(_angle); float ret = _ret / (float)TRIG_MAX_RATIO; return ret; }

Turns out it's drawing a line that only needed a one-line helper.

static void canvas_draw_line(GContext *ctx, int x1, int y1, int x2, int y2) { graphics_draw_line(ctx, GPoint(x1, y1), GPoint(x2, y2)); }

Here's the arc drawing helper.

static void canvas_draw_arc(GContext *ctx, GRect rect, int32_t angle_start, int32_t angle_sweep) { angle_start += 90; GOvalScaleMode scaleMode = GOvalScaleModeFitCircle; int32_t angle_end = angle_start + angle_sweep; graphics_draw_arc(ctx, rect, scaleMode, DEG_TO_TRIGANGLE(angle_start), DEG_TO_TRIGANGLE(angle_end)); }

I guess Android and pebble define 0 degrees at a different position on the screen, hence the adding of 90. And pebble wants an angle end, not a number of degrees to extend the arc. And there's those weird pebble degree units again.

[a1291762]

In a pebble app, you can include math.h and invoke sin, but it crashes with some values. Turns out you're supposed to use their special sin function, which is cool and all, but now my ported-from-Android drawing code needs a helper function. Ditto for cos.

Pebble rects use the sane x y w h rather than Androids insane x1 y1 x2 y2.

Drawing an arc is a little different to Android but a one-line helper function makes it the same. But strokes have rounded ends, not straight ends.

[a1291762]

I doubt the whole thing will squish down to the pebble but this is a start...

[a1291762]

It was only a few days ago, when I saw an ad for a documentary that I finally connected 'stan' with the Eminem song.

I've understood stan, to stan something, to be a stan, etc. But I somehow missed the origin of the term.