Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Feb 13 19:03
    lassik commented #410
  • Feb 11 18:08
    gassechen commented #485
  • Feb 11 17:34
    feeley commented #485
  • Feb 11 15:37
    gassechen commented #485
  • Feb 11 14:53
    gassechen commented #485
  • Feb 10 22:03
    gassechen commented #485
  • Feb 10 20:49
    feeley commented #485
  • Feb 10 19:30
    gassechen commented #485
  • Feb 10 19:17
    feeley commented #485
  • Feb 10 18:44
    gassechen opened #485
  • Feb 08 23:00
    Travis gambit/gambit (master) passed (1377)
  • Feb 08 22:23

    feeley on doc-fix

    (compare)

  • Feb 08 22:22

    feeley on master

    Fix missing line in example Merge pull request #484 from ga… (compare)

  • Feb 08 22:22
    feeley closed #484
  • Feb 08 21:58
    Travis gambit/gambit#484 passed (1376)
  • Feb 08 21:21
    Travis gambit/gambit@2ee3f6e (feeley/doc-fix) passed (1375)
  • Feb 08 21:18
    Travis gambit/gambit (master) passed (1374)
  • Feb 08 20:41
    feeley opened #484
  • Feb 08 20:41

    feeley on doc-fix

    Fix missing line in example (compare)

  • Feb 08 20:38

    feeley on openssl-eof-fix

    (compare)

Randall Gray
@snarkypenguin
I'm trying to make an interface to the Proj library for handling geographic coordinates, but I am having trouble with "const" parameters ... usually things like constant character strings, but sometimes more exotic. Is there a "const" indicator I can use when defining C types?
I plan on a netCDF interface too, but proj seemed easier to start with ;-)
Marc Feeley
@feeley
@snarkypenguin unfortunately const is not supported by the C interface... it has been on my todo for a while. most C compilers have a switch to not treat the absence of const as an error, so this might be a workable temporary measure
Marc Feeley
@feeley
@amirouche yes I'll reboot the server later today (actually turned it off yesterday due to overheating and forgot to turn it back on)
Randall Gray
@snarkypenguin
Thankyou, I didn't even think of that (randall smack his forehead foolishly). When it is working and tested, I'd like to contribute it to the dumping ground.
Marc Feeley
@feeley
@snarkypenguin I'm sure a better place than the dumping ground can be found for your module. more on this later...
Marc-André Bélanger
@belmarca
@feeley Are you going to release a package manager?
:)
Marc Feeley
@feeley
@belmarca would you use it if there was one?
Marc-André Bélanger
@belmarca
yes if that means more code sharing :)
Marc Feeley
@feeley
more on this will be presented at the "Gambit at 30" meeting... for details see https://mailman.iro.umontreal.ca/pipermail/gambit-list/2019-July/009117.html
amirouche
@amirouche
yeah! Eager to read more about all those new stuff!
Marc-André Bélanger
@belmarca
are there any presentations scheduled for Gambit at 30?
Marc Feeley
@feeley
yes... and maybe one of them is from you!
Marc-André Bélanger
@belmarca
I suppose that could make the other presentations that much more interesting :laughing:
you can contact me if you are looking for presenters
Marc Feeley
@feeley
I am looking for presenters... you have been contacted... your move...
Marc-André Bélanger
@belmarca
Ok then
Marc Feeley
@feeley
there will be 20-ish participants, so it is possible (and it would be nice) if everyone has something to present
Marc-André Bélanger
@belmarca
OK, if you have some place to register the subject that could eliminate possible duplicates.
Marc-André Bélanger
@belmarca
something related that could help is publishing a roadmap or to-do list, other than the current github issues.
amirouche
@amirouche
the issues are good enough, maybe add some issues for the existing task with a help needed flag
I started working on the JavaScript backend, and keep updating the issue 404
The roadmap could be announced on the mailling list tho
Marc-André Bélanger
@belmarca
well for example I see there is issue #408
That one mentions the missing syntax-rules, however I wouldn't necessarily go looking for that information inside the r7rs issue
there is one project but it is dated 2017
amirouche
@amirouche
well that is true
maybe we should go through the whole R7RS and create table and flag features that are done and those that are missing
Marc-André Bélanger
@belmarca
from what I understand there should be discussion about this in october
amirouche
@amirouche
FWIW, I am willing to move to use gambit both in the frontend and backend...
even if the backend requires multi threads and gambit smp support is new
what I holding me back, was mostly the lack of R7RS library system
Marc-André Bélanger
@belmarca
why do you absolutely need smp?
amirouche
@amirouche
I build a database on top of wiredtiger which is embedded key-value store library, so I need queries not to block other queries.
Bradley Lucier
@gambiteer
@amirouche I did something like this using threads, but no smp.
Does using wiredtiger mean you need smp?
amirouche
@amirouche
@gambiteer what is the difference between smp and threads? I was under the impression that thread in gambit were different from POSIX threads.
I need that one "thread" doesn't block other threads, as concurrency issues are dealt with by wiredtiger.
Marc Feeley
@feeley
@amirouche where is the blocking happening? if it is on a file descriptor then Gambit's green threads will be sufficient because the thread scheduler multiplexes the CPU so that Scheme threads waiting on a fd are put to sleep without blocking other threads. Scheme-level mutexes and condition variables also don't block other threads than the one blocking.
amirouche
@amirouche
the blocking call happen on ffi call
the main resource used is cpu
Marc Feeley
@feeley
I'm not sure what you mean with that. I'm asking to know what the OS thread is blocking on (reading a fd, locking a mutex, etc). I don't know how wiredtiger is implemented.
amirouche
@amirouche
wiredtiger is implemented in C, the scheme code is blocking on C calls
Marc Feeley
@feeley
in that case, a possible approach is to allocate an OS thread with each Scheme thread, and have the OS thread do the calls to wiredtiger, and to communicate with the OS thread through a pipe... it is not as complex as it sounds
amirouche
@amirouche
IIUC a pipe is two ports?
Marc Feeley
@feeley
I mean an OS pipe (which has two ends). Ask @jazzscheme who has used this for streaming audio/video with separate OS threads.
Guillaume Cartier
@jazzscheme
My first implementation was as Marc said to spawn an OS thread (a pthread in my case) and have the thread communicate “completion” by sending 1 byte down a named pipe. The named pipe is only read on the Gambit side where the Gambit thread will be blocked and only written on the C side. Later on, when porting to Windows where named pipes are a bit different and touchy to use, I tried replacing the named pipes by TCP sockets and tests showed performance to be identical as the OS clearly optimizes the special case of same machine (in this case even same process) communication. All this works very well for my needs and with it I can now play 80 movies at the same time on my mac (160 Gambit threads blocked on receiving codec data).
amirouche
@amirouche
tx
Phiroc
@Phiroc
Hi Marc, I've noticed a few bugs in Not Emacs on iOS. Among others, there's no way to close the initial Help Window, you can't quit Scheme (pressing Ctrl-D several times doesn't work), and "yank from operating system" displays a window into which you can't paste text from other iOS applications. Furthermore, being able to open files from iCloud or DropBox would be a bonus. If I were a C/iOS programmer, I could probably fix those bugs myself, but I'm not. Can you help? Many thanks. Cheers, P