dylan-lang/general

Dylan programming language -- http://opendylan.org -- http://play.opendylan.org

promovicz

@promovicz

laughs

promovicz

@promovicz

I've made some progress with Marlais. With refactoring-heavy jobs like this I tend to keep backlog and squash it a little before pushing.

Carl Gay

@cgay

@Slids :point_up: January 17, 2020 12:41 AM

Jon Godbout

@Slids

tanks!

Carl Gay

@cgay

Oh man, link-target needs some love.

(file-system)

I'm going to document it but not going to shave this yak just yet.

Peter S. Housel

@housel

What's wrong with it?

Seems straightforward

(Other than outdated Windows support)

Carl Gay

@cgay

It looks like if this unless https://github.com/dylan-lang/opendylan/blob/master/sources/system/file-system/unix-file-system.dylan#L162 does not call unix-file-error the function will infinite loop because link won't change. ?

Peter S. Housel

@housel

Ah. Yes, that should get fixed

Carl Gay

@cgay

But there are no tests so ... :)

It's just that feeling of "oh, it's not documented. No worries I'll just quickly write some simple doc for it. Oh, it's not tested, and oh isn't that a bug?"

Peter S. Housel

@housel

Indeed

Carl Gay

@cgay

TIL that when I put TODO(b/12345): in my code, the TODO shows up in the bug tracker for b/12345. Boy do I love good tooling. (Also, when I close b/12345 I'll get an email at some point reminding me to remove the TODO.)

Peter S. Housel

@housel

I’ve finally got my debugger-nub back to the point I had it at in late July, before I started a major refactor to split it into separate CORBA and debugger domains (and spent some time adding bits of functionality to the LLVM project's JIT support, and had to adapt to various evolving JIT interfaces)

promovicz @promovicz listens with interest. I've been wondering about CORBA turning up in the changelogs.

promovicz

@promovicz

I've had a slight problem with dockerization on debian buster. opendylan/configure doesn't seem to deal with llvm/clang versions too nicely yet. I'll figure that out later though. I'm at the wrong set of machines right now.

I believe my (booted with release) opendylan thinks it should use the (batteries) LLVM that was included with the bootstrap.

the system version might be way-old though. I guess LLVM distribution is going to remain a pain. I recently talked to a FreeBSD packager who listed me around ~15 LLVM build trees in ports.

packagers hate "bring your own" versions of course. but we wouldn't be so bad because we don't require constant rebuilds like rust does. which for maintainers might mean a multiple-day wait for the ARM and risc-V machines.

Peter S. Housel

@housel

On my RISC-V machine a full LLVM rebuild took about three-quarters of a day

Peter S. Housel

@housel

I looked at a couple of official Docker language images and they don't create users to run things as, they just do things as root

Carl Gay

@cgay

My lack of thinking like a C programmer has struck again. "Each index MUST be in the range [0, "charcnt" - 1]; it designates the NUL-terminated string of octets starting at position "idx" in the time zone designations." I was not expecting a single run of non-zero characters to represent two strings, one of which is a suffix. :)

Jon Godbout

@Slids

Is there a way to get the bits representing a thing:
Take
let foo = 'c';
I want to see the bit representation for foo and compare it to the bit representation for as(<integer>,'c')

Peter S. Housel

@housel

Not within the standardized language; there are some implementation-specific primitives that can be used to do it

For that particular case, characters like 'c' have low-order tag bits of #b10, and integers have low-order tag bits of #b01

Carl Gay

@cgay

I like a lot of https://flix.dev/principles/ but I think not supporting binary literals is going a bit too far. (I could do without octal though.)

Jon Godbout

@Slids

thanks, always nice to be able to boil things down to the bits sometimes

promovicz

@promovicz

@Slids dylan is very object-oriented. it hides representation. you can use it like a scripting language in that sense. just let the compiler worry about it.

the advantage is that you can sort-of treat all things equal. everything is an object. everything is the instance of a class.

Jon Godbout

@Slids

I understand, but I want to know

(Even though it doesn't matter)

It's interesting the hacks I've made when I don't know how to do something. I was trying to convert the string

"123"

into the vector

[1, 2 ,3]

and didn't see the as operator...

promovicz

@promovicz

oh, sure. there are ways to figure out how things get represented by dumping IR. as far as objects go there are some docs if you want to know the details.

I'm not sure if there is a quick off-hand inspector specifically for that.

promovicz

@promovicz

AFAIK a <byte-character> like that is an immediate value of some sort. pretty baked in. you can expect that to be as fast as any other small integer.

and if you declare slots on it you will also get efficient slots.

Jon Godbout

@Slids

declare slots?

Peter S. Housel

@housel

That’s not usually the case with Open Dylan; most slots use the regular <object> tagged representation

promovicz

@promovicz

well, if you declare slots as <byte-character> or <integer> then these objects will not be boxed unless required.

Peter S. Housel

@housel

That’s true with Gwydion, but not Open Dylan

Carl Gay

@cgay

I thought that's what repeated slots were about. ?

Peter S. Housel

@housel

Objects can hold raw values, and repeated slots of raw values can be smaller than word-sized

promovicz @promovicz yields to housel because he will know these things better. my memory of those things is a bit fuzzy after to many language implementations I guess.

Carl Gay

@cgay

Greenwich = UCT = Universal = UTC = Zulu

Peter S. Housel

@housel

Well, actually… 😀

Carl Gay

@cgay

These are links that all point to the same zoneinfo file.

Where communities thrive

People

Repo info

Activity