Tom Scogland

[Tom Scogland @tscogland]

@oPryzeLP I have a question regarding the Universal Minecraft Tool... it looks like I’m getting some poplar variants converting to pale oak, specifically the leaves, is there a way I can put in a custom rule or something to take care of that? Everything else is coming out great.

[Flux @FluxFramework]

new releases for flux-core, flux-sched, and flux-accounting are out now! 🎉 https://t.co/ktsnJnbquL

highlights:

- support for CLI plugins
- syslog capability for the broker
- new "flux-content checkpoints" command
- add enforcement of a per-queue max nodes limit

[Tom Scogland @tscogland]

@__phantomderp Summary looks nice. Are you planning to have a way to get one of these as a function pointer and opaque pointer as well? I use that pattern with c++ and rust frequently, but since c has no way to do templates/generics I wouldn’t be able to as a user. It has to be language level.

[Tom Scogland @tscogland]

@__phantomderp How would you like feedback? I worked on one of the previous papers you mention, and would be willing to help hash some things out. Generally I agree with where you’re going FWIW, though I’m curious how you intend to handle the finalizer issue for captured entities.

[Tom Scogland @tscogland]

@__phantomderp @peter0x44 Is it not applying early enough? I’ve run into issues with the auto detection commands cmake uses to determine the libc location and so forth not following variables like that. Frustrating to be sure, but can be worked around with a wrapper or similar if that’s what it is.

[Tom Scogland @tscogland]

@__phantomderp bfd’s just not optimized for the way windows filesystems and fork and so-forth behave. The operations it relies on being fast just aren’t. I haven’t done this in a while, but gold at least used to work in mingw land, parallel too. lld also works with recent gc with -fuse-ld=lld

[Tom Scogland @tscogland]

@joewantsadrink @bbsourpatch I’ve wondered about this, if it’s in the mode to unlock only when the user is focusing on the phone (eyes open, on the phone, etc.) can they compel you to do so? I that your likeness is not testimony, but it is compelling positive action that could be self incriminating.

[Tom Scogland @tscogland]

@__phantomderp It is pretty funny as a gap. Windows has no way to jump a file picker window to a dragged file, or access the file or folder of the current open file of an arbitrary application window. Mac has had that one for 20+ years. Also no native PDF handling. It’s all trade-offs. 🤷‍♂️

[Tom Scogland @tscogland]

@__phantomderp @edefic This bit, right here, is what drives me nuts. u?int\d*_t should not be optional. The fast and least variants are something, granted, but the fact the actual specific width types aren’t required is extremely painful. (Just had to write a bunch for OpenMP’s spec due to the lack)

[Tom Scogland @tscogland]

@ericniebler @Nekrolm As awful as it is, I now wonder if doing “void (*b)(S) = [](S s){ baz(s); };” would optimize out to get to the same place or always result in the extra call. Seems pretty arbitrary.

[Tom Scogland @tscogland]

@vsoch @ubuntu It’s super sad, but this and snaps and breakage is why I’m giving up on Ubuntu. I’ve been a die hard user since I didn’t have time to manage my gentoo fleet anymore, but I just can’t do it. Back to Debian, gentoo, maybe after the fully reproducible container arch. 😞

[Tom Scogland @tscogland]

Pleasant surprise of the day, code written in 2014 in PyOpenCL of all things, just ran today with zero modifications on brand new platform, new hardware, new instruction set. Much as it was a pain, there's something to that kind of portability.

[Tom Scogland @tscogland]

@tgamblin Day, decade, determinism, so many things that fit if it starts with lost 😂

[Tom Scogland @tscogland]

@tgamblin @ronnybrendel TBH I think it usually is. On a Mac, it’s trivial in anything but office. How do you get the containing folder of a document in any windows program to open in explorer? KDE app? Gnome app? It consistently shocks me how oddly difficult it is to do this kinda everywhere but aqua.

[Tom Scogland @tscogland]

@brockpalen @tgamblin @vsoch @FluxFramework And as @grondo@github says: And since Flux can run under any RM, your stuff will work anywhere
And it's true. Run flux under whatever (bate metal to mpirun) and your scripts are portable nearly anywhere, including clusters, Raspberry Pis, and, at least once, a refrigerator.

[Tom Scogland @tscogland]

@brockpalen @tgamblin @vsoch @FluxFramework The problem is people having to write custom schedulers to manage their RM. We have APIs in C,Python and lua, non-viral licensing, and a highly configurable and replaceable scheduler. Use our API to submit the pieces and let us schedule it, stop writing schedulers! It's hard!

[Tom Scogland @tscogland]

@vsoch FHS makes me sad. Not because things have known relative placement, but the assumption that everything must be merged

[Tom Scogland @tscogland]

@vsoch @FluxFramework First resource management, then THE WORLD!

[Tom Scogland @tscogland]

@fmzakari Wait till you see the extensions to support GPUs. Dynamically sized register sets and dynamic offsets for live values oh my https://t.co/ewArUx0KOY Mind boggling. Regular gdb live-stepping code on my GPU though? Epic.