Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Jan 29 22:26

    cgay on main

    README.md: switch RST markup to… (compare)

  • Jan 29 22:24

    cgay on main

    Initial commit Initial commit -- just bare bon… First pass - generates index.rst and 16 more (compare)

  • Jan 29 22:17

    cgay on master

    doc: Fix invalid RST topic refe… (compare)

  • Jan 29 22:09

    cgay on master

    doc: Minor RST markup fix (compare)

  • Jan 29 21:29

    cgay on master

    Run channels tests as part of `… channels-test-suite: use descri… Merge pull request #1476 from c… (compare)

  • Jan 29 21:29
    cgay closed #1476
  • Jan 29 18:27
    cgay opened #1476
  • Jan 29 02:53
    cgay synchronize #1474
  • Jan 29 02:28

    cgay on master

    dylan-tool 0.8.0 Merge pull request #17 from cga… (compare)

  • Jan 29 02:28
    cgay closed #17
  • Jan 29 02:19
    cgay opened #17
  • Jan 29 02:17

    cgay on v0.8.0

    (compare)

  • Jan 29 01:57

    cgay on master

    dylan build: prepend workspace … dylan publish: Specify catalog … Merge pull request #52 from cga… (compare)

  • Jan 29 01:57
    cgay closed #52
  • Jan 28 04:45
    cgay opened #52
  • Jan 26 19:51
    cgay opened #14
  • Jan 25 12:31
    Stasenko-Konstantin starred dylan-lang/opendylan
  • Jan 22 17:11
    cgay synchronize #1474
  • Jan 22 06:14
    cgay opened #51
  • Jan 22 05:45
    cgay ready_for_review #1474
promovicz
@promovicz
one thing that I really like about the direction this interpreter is going is the ability to define any integer type as a limited integer type. I should be able to represent any C integer, and that should help a lot with FFI.
not that c-ffi ain't any good. I actually think it's excellent.
(within its limitations)
for this sort of thing it's actually an advantage that our base language is small compared to the libraries.
marlais can definitely be a complete dylan. that wouldn't be so easy if we had more subsystems in the language core.
promovicz
@promovicz
with regards to macros I'm still thinking. I'm tending towards wanting a macro expander implemented in dylan because we already have that code. the parsers of the interpreters would have to emit a (possibly shared) d-expr representation.
that way the code could be borrowed from dfmc, dbmc or d2c. they all kind of do this already. the dfmc version probably has the code with the best factoring.
it also has the best features (template calls).
Jon Godbout
@Slids
How do I get format statements to print before dylan finishes?
Carl Gay
@cgay
force-out() and/or force-err()
I would like these to happen by default, but I think maybe that's expensive? Maybe we could set them up to happen by default (line buffered?) but have a way to turn them off for performance. I think we discussed this at one point...I'll check history.
Carl Gay
@cgay
Only this :point_up: March 21, 2019 12:42 PM
Is the new Dune movie any good? Should I use my $10 Google Fi credit on it?
Peter S. Housel
@housel
I have line buffering half-implemented in a stash someplace
(Of course some of my stashes are named things like with-many-snerk and moths-and-butterflies so I may or may not be able to find it)
Jon Godbout
@Slids
the new dune movie is only part 1 of n
Not defaulting to outputting the values hurts the printf(format-out) debugging I always do... Any chance of a repl? :)
Peter S. Housel
@housel
Soon
(Hopefully by mid-Spring)
promovicz
@promovicz
@cgay stdio in libc makes stderr unbuffered. that's how they deal with it. they also flush atexit IIRC.
a working repl in opendylan might be a magnet for me. :)
Carl Gay
@cgay
format-outf == format-out(...) + force-out() :smirk:
Only 1/2 joking. I've been using a debug-out that is basically that.
Carl Gay
@cgay
Flushing at exit would solve the biggest part of the problem. It's one thing to not see the output right away and it's another thing altogether to never see it at all.
Peter S. Housel
@housel
Everything does get flushed at exit
(unhandled exception aborts possibly excepted)
Carl Gay
@cgay
hmm, I could swear I've seen that not happen. yeah, maybe due to errors.
Carl Gay
@cgay
I just noticed that showing macro expansion fragments (or whatever the right term for ?"expr1" is) shows the associativity of infix operators. At least I think that's what's going on when testworks turns my 5 * 60 * 60 into: ((5 * 60) * 60) = zone-offset-seconds(us-eastern, time:, t1) failed [18000 and -18000 are not =.]
Carl Gay
@cgay
I thought if I have /usr/lib/x86_64-linux-gnu/libzip.so.4.0 and I put c-libraries: -lzip in my LID file I should be able to link against this, but I'm getting Internal error: Link failed: Link failed: /usr/bin/ld: cannot find -lzip.
Ah, it fails for the zlib library as well, so I guess there is probably something misconfigured on my system.
Carl Gay
@cgay
Linking libzip.so to libzip.so.4.0 works, but sadly I have no idea why I should be required to do that.
Peter S. Housel
@housel
You don’t have libzip-dev installed
You should be able to run pkg-config --libs libzip and get the right link invocation
Carl Gay
@cgay
ok, thanks. But I guess my bigger problem is that I don't understand the general philosophy. Or maybe that's too grandiose of a word; I don't understand why, if simply creating a link is enough, they don't just do it for all installed libraries in advance...
It helps to read the overview for pkg-config (which was apparently written by a guy named Havoc). I'll try to remember that's what I need in cases like this.
Peter S. Housel
@housel
Particular programs are set to runtime-link against libzip.so.4.0, but the linker doesn’t search through all the possible .n.m versions in its link path
The dev package provides the symlink .so that the linker will find
But if you’re just running programs that need libzip but not developing against it then you don’t need that
Carl Gay
@cgay
So at link time presumably the linker resolves the symlink and embeds the specific version in the executable?
Or more likely it has nothing to do with the file name but is embedded in the .so itself, I suppose.
promovicz
@promovicz
Making small strides in marlais... the code is now warning-free and I'm starting to work on the interpreter core.
It is also pretty namespace-clean now. I'm currently working to collect the interpreter state into a context structure, among other things.
its definitely fun to do this with interaction. it feels natural to dylan.
It seems to me that marlais can be "compatible without macros" in a few more months.
My goal definitely is library reuse.
promovicz
@promovicz
Macros are important to some libraries but not to others. I haven't estimated yet how much work real macro support would take...
I still prefer marlais over mindy because it is so much simpler. It's the ideal testbed.
Carl Gay
@cgay
Sounds like good progress.
I'm working on getting rid of shelling out to git in pacman and download via HTTP instead. Hence the work on libzip.