Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Curtis Rueden
    @ctrueden
    But in the meantime, you set it to non-strict, you won't see those traces and the context will likely get created (mostly) successfully.
    It might be missing services though, which may lead to bad behavior that is less fail-fast than the current behavior. So it's your call.
    Ulrik Günther
    @skalarproduktraum
    @ctrueden i am poking my head again into the scripting-kotlin thing. out of sheer madness. it seems that most of the issues that have been plagueing us are fixed in kotlin 1.3.50, however there still seems to be a problem with connecting services and i/o parameters. can you tell me where they are actually handed over to the ScriptEngine’s bindings?
    Philipp Hanslovsky
    @hanslovsky
    @skalarproduktraum Is one of the issues problems with exceptions: Whenever evaluation throws an exception, the interpreter gets stuck in a bad state and all subsequent evaluations throw exceptions as well. I did not dig into it, though, and did not test with 1.3.50, either.
    Philipp Hanslovsky
    @hanslovsky
    Unexpected behavior with @Plugin annotations in Kotlin files in src/main/java instead of src/main/kotlin: https://forum.image.sc/t/plugins-annotations-are-not-added-to-index-in-jar-file-if-implemented-in-kotlin-in-src-main-java-source-tree/29700
    Curtis Rueden
    @ctrueden
    @hanslovsky Shouldn't Kotlin files go in src/main/kotlin though?
    Philipp Hanslovsky
    @hanslovsky
    No need to, they can be in src/main/java, too.
    Ulrik Günther
    @skalarproduktraum
    @ctrueden is there an “easy” way to fix the java 11 issue mentioned in https://github.com/scenerygraphics/sciview/issues/222#issuecomment-533909217 or is this a horrible, multi-headed beast? if you have a pointer what’s the origin there, i’m more than happy to contribute sth
    Curtis Rueden
    @ctrueden
    @skalarproduktraum Are you talking about the library linkage issue?
    Without digging deeply, I was thinking it should be straightforward to improve the imagej-launcher to look in more locations.
    Stefan Helfrich
    @stelfrich
    @ctrueden @skalarproduktraum I am not entirely sure but the (unreleased?) master of imagej-launcher actually adds checks for default install locations of OpenJDK on Linux
    At least I remember working on this...
    Ulrik Günther
    @skalarproduktraum
    @ctrueden @stelfrich that’s helpful info, thanks :-) now i am wondering what is actually the latest released imagej-launcher version, because the log in the linked issue also includes a message about imagej-launcher not having the XInitThreads fix i committed a while ago … has there been a release recently? otherwise, this is still weird because then the sciview update site should ship the launcher with the fix (maybe @kephale know something, too? :-D)
    Stefan Helfrich
    @stelfrich
    @ctrueden and myself promise to follow up on this … soonish … we are focussing!
    Curtis Rueden
    @ctrueden
    "Soonish" means maybe tomorrow. Writing paper. Must... not... do technical things now...
    Ulrik Günther
    @skalarproduktraum
    oooh paper, what are you guys writing?
    Kyle I S Harrington
    @kephale
    @skalarproduktraum iirc 5.0.2 was cut by @ctrueden prompted by the need for the xinitthreads
    the code is there in imagej.c in that release
    5.0.1 was a wonked travis build
    Ulrik Günther
    @skalarproduktraum
    https://github.com/imagej/imagej-launcher still says 4.0.5 though o_O
    and 4.0.5 is also what’s in the log
    imagej-launcher version smaller or equal to non-working version (4.0.5 vs. 4.0.5), disabling Vulkan as rendering backend. Disable check by setting 'scenery.DisableLauncherVersionCheck' system property to ‘true'
    Kyle I S Harrington
    @kephale
    on central
    Ulrik Günther
    @skalarproduktraum
    ah, sure, thanks!
    that still leaves the question why there’s 4.0.5 in a log from a few days back
    Kyle I S Harrington
    @kephale
    yeah, i think i've seen that as well, im not clear on whether the updated launcher is shipped with everything yet
    Ulrik Günther
    @skalarproduktraum
    but then the launcher from the sciview site should take precedence, not?
    Kyle I S Harrington
    @kephale
    yes
    Kyle I S Harrington
    @kephale
    is there a way to control the name of an object in ObjectService. Specifically, I want to control the names that show up in drop down lists for
    @Parameter
    Node n;
    Jan Eglinger
    @imagejan
    @kephale the WidgetModel is currently taking the object's toString():
    https://github.com/scijava/scijava-common/blob/097cbb4b4248b84f3ad475b5a844fe3cc8b1ca97/src/main/java/org/scijava/widget/DefaultWidgetModel.java#L235-L242
    So you might have to override toString() for your Nodes, or we change the model to ask for getName() or some such.
    Kyle I S Harrington
    @kephale
    @imagejan <3 thank you!
    NicoKiaru
    @NicoKiaru
    Yep, and it's pretty annoying when you can't override it. I'm dealing with this all the time. @imagejan do you think there's a way to change this behaviour (and potentially without wrapping all objects into a renamable one) ?
    Jan Eglinger
    @imagejan
    Yes, submit a pull request to scijava-common or scijava-ui-swing... I'm unsure where exactly this has to be changed. We probably need to dig a bit.
    Curtis Rueden
    @ctrueden
    Regarding object names... we could change the DefaultWidgetModel to first look if the object implements a Named interface and if so, call getName() rather than toString(). Would that help?
    Kyle I S Harrington
    @kephale
    although i can override toString() in this case, i think @NicoKiaru wants to do something more like passing a lambda that could be called on an object of a given type to get a name
    but i dont actually know his use case
    Curtis Rueden
    @ctrueden
    We could change the ObjectService to have new API that lets you add an object with an explicit name, sure.
    But by "we" I mean "someone in the community files a PR"
    Kyle I S Harrington
    @kephale
    :P
    Curtis Rueden
    @ctrueden
    You'd have to study the ObjectIndex data structure carefully and think about how best to achieve it.
    But in principle I'd be open to an enhancement like that.
    I think using a lambda per type is overkill, though.
    Heh, I should just shut up and let @imagejan's answers stand—he already said all the things I said. :stuck_out_tongue_winking_eye:
    Kyle I S Harrington
    @kephale
    i think we're ok on the sciview side at least
    just needed some changes in scenery
    NicoKiaru
    @NicoKiaru
    thanks for the directions everybody! FYI my use cases are currently passing objects like BdvHandle, Transforms and SpimData objects in `Commands.
    Ulrik Günther
    @skalarproduktraum
    hi everyone! i just bumped to the latest scijava pom on cleargl, and am now getting the error Error assembling JAR: Invalid automatic module name: '' -> [Help 1] — any hints what i’m missing?
    Curtis Rueden
    @ctrueden
    @skalarproduktraum This bug is fixed for pom-scijava on master, so pom-scijava 28 will not have this bug anymore. But in the meantime: declare a <package-name> property with your package prefix. (This is a good idea anyway.)
    Ulrik Günther
    @skalarproduktraum
    Will do, thanks @ctrueden