I’m seeing udp-receive! lose packets. It happens approximately the same time as a major GC, so I think they’re related, but I’m not sure. Does that sound familiar to anyone who’s done this?

I understand UDP is inherently unreliable. If Racket didn’t drop the packets, something else might. So on the one hand, this is fine — it’s helpfully exercising my loss-detection. My sender retransmits and eventually all is fine. On the other hand, it does add somewhat to the time-to-success. And I’m just curious if udp-receive! is intended to be unreliable during GC, this way.

(To clarify, I’m seeing this happen with the sender and receiver as two Racket processes bound to different UDP ports on the same machine. No interweb tubes involved.)

(To clarify #2, these are small 1200 byte datagrams.)

That was on 6.10. I see similar on 7.0.

I just downloaded 7.1.0.10 CS for macOS, and I’m getting an error. Making minimal example….

Running something like this: #lang racket/base (require racket/udp racket/function) (define port 50000) (define u (udp-open-socket #f #f)) (udp-bind! u #f port #t) (thread (thunk (define buffer (make-bytes 1200)) (for ([(len host port) (in-producer (thunk (udp-receive! u buffer)))]) (displayln (subbytes buffer 0 len))))) (sleep 2) (udp-send-to u "127.0.0.1" port #"Hello") (sync never-evt)

Outputs: $ /Applications/Racket_v7.1.0.10-CS/bin/racket /tmp/udp.rkt #<thread> invalid memory reference. Some debugging context lost context...: condition->exn dynamic-wind winder-dummy /Users/racket/build/plt/build/cs/racket/src/build/cs/compiled/io.scm:37450:3: do-udp-receive!28 /tmp/udp.rkt:12:2: for-loop /tmp/udp.rkt:10:1 internal error: terminated in atomic mode! context...: /Users/racket/build/plt/build/cs/racket/src/build/cs/compiled/thread.scm:1228:2: internal-error /Users/racket/build/plt/build/cs/racket/src/build/cs/compiled/thread.scm:10114:18 invoke

It’s probably a simple fix on line 37,450 of io.scm? :smile:

Simplified a bit more and added as https://github.com/racket/racket/issues/2433

11

The number of questions on MetaPost on StackOverflow.

Does that mean that MetaPost is “dead”?

probably

I don’t know any TeX/LaTeX users who use metapost

@soegaard2 I am downvoting because you should discuss this on meta.meta.stackexchange.

meta meta !?

a few years back it was pstricks, nowadays almost everyone (including myself) is using tikz/pgf

This recursion has no base case and will end badly. :slightly_smiling_face:

I wonder if Knuth uses metapost

Found the Metapost mailing list. Roughly one thread a month.

What made tikz take over?

even metafont seems to be going away? I think with the truetype engines it becomes unnecessary, but I don’t know, maybe it will still be necessary

Ah y’all mean https://stackoverflow.com/questions/tagged/metapost not https://meta.stackexchange.com/ Welp. Never mind. ¯_(ツ)_/¯

tikz has a lot of functionality and you can work in many levels, from the level of basic shapes and forms to very high level constructs like “these nodes are a tree, draw it for me”

and it’s independent of the backend, pstricks was tied to postscript generation

There is an impressive number of ready-made packages for tikz.

yeah