@robby Basically, I would like to take the definitions text% field in DrRacket and place it in a new window, if that’s possible.

@leif the mixins in macro-debugger/syntax-browser/text might be useful

@ryanc Oh cool, thanks. I’ll take a look at that.

Cmd-n isn’t what you want, I presume, but it does that job.

(So I am still not getting it.)

Hmm…well you know how you can use the slideshow/repl package to embed a racket:text% object in a slideshow, I would like to put a text object in a slideshow and still have arrows being drawn.

Does that make any more sense?

Although as a side note (and @mflatt might know this one), in slideshow/repl, is there any way to have interactions typed into the result-area take place inside of a module, like the interactions area in drracket?

Or at least have the provided variables bound?

@leif That might be possible by supplying make-namespace functions to repl-area and make-repl-group so that they work together, but I wouldn’t be surprised if an extra hook is needed to make it work well.

hello @mflatt I was using PLTADDONDIR environment variable to configure a private addon directory to install various packages for a Racket application’s use. I noticed that if I give relative paths to the different addon directory I get collection not found errors. It appears the addon directory is being resolved relative to location of the installed package instead of the working directory I was in when I export the PLTADDONDIR variable which I had naively been expecting. If I use absolute paths, it works fine which isn’t a surprise. Documentation doesn’t mention addon-dir should be given as absolute path, is this something that should be mentioned?

@pocmatos I’ve used scribble/srcdoc a little (in Frog) and I like it for reference docs. (I haven’t gone back and rewritten existing docs to use it, but I’d use it for new projects.) Of course I’d still have a main container .scrbl for prose and @include-extracted each .rkt with a srcdoc mod. I did find I wanted a little sugar around proc-doc such as https://github.com/greghendershott/frog/blob/master/frog/private/define-doc.rkt.

@abmclin The PLTADDONDIR environment variable is converted to an absolute path when Racket reads the variable on startup, so normally only an absolute path makes sense. I’ll update the docs to clarify.

@mflatt I’d like to try to get the Travis and AppVeyor builds working again before merging racket7 — any thoughts on either the db+macOS issue on Travis, or the strdup issue on Windows (my attempted fix for the strdup issue didn’t work)?

@ryanc How well it work for the db concurrency test to use (sync (system-idle-evt)) instead of sleeping and measuring time? Your comment before characterized the test as checking whether other threads get to run, and (system-idle-evt) is a general way to test that property (as long as you can assume no other tests running in threads in the same place).

@samth Probably “msvcrt.dll” needs to be specified to get “strdup” or "_strdup" on Windows. I’ll see if that makes it work.

@mflatt great! why did it produce that strange crash with #f?

I’m used to getting “undefined symbol” for issues like that

I guess it’s not so much that “mvcrt.dll” is needed as that you need to use a _strdup and free that connect to the same allocator. (Only on Windows is there a question of which free you use.)

ah, that makes sense

DrRacket (6.12) on Windows brings unknown certificate message upon startup

Which version of Windows? (I can test 8 &10)

windows 10

Ok I’ll try

@mflatt I’m not sure how to use that in a test. To recap the problem: suppose T1 and T2 are threads, and T1 executes a db operation; I want to check that T2 is scheduled periodically for the entire duration of T1’s db op. If I only perform a single check from T2, then I can’t be sure that T2 didn’t sneak in just before or after the db operation. So I could have T2 test (sync/timeout 0 (system-idle-evt)) over and over, but that doesn’t seem better than the sleep and counter approach. I’ll change the test’s fudge factor (meant to earlier, but forgot).

If T2 can’t get past (sync (system-idle-evt)) by “sneaking in”. T1 would have to be specifically blocked somehow – on a semaphore, file descriptor, or something else that the scheduler would know about.

Oh, I see, right. It’s not exactly the property I want, but it sounds like a reasonable proxy. I’ll try that instead.