Clojurians Log v2

I can't believe it took me this long, but I finally got around to logging the question @alexmiller @dmiller : https://ask.clojure.org/index.php/13072/feature-request-clojure-specify-conditional-values-local

seems like a good consideration for tools.deps if you want to make a jira in TDEPS or a question on http://ask.clojure.org

@alexmiller OK. The issue may be of wider interest than just Windows running Linux. I know that when running Parallels on Mac there is similar "integration" between the file systems. So perhaps this can be framed as a more general problem, one that devs today are pretty likely to run into.

At least, I ran into it. FWIW. lol.

certainly the code in tools.deps is deferring to the jdk as much as possible for path and classpath joining, splitting, etc so that it is os agnostic in that regard. it would be helpful to have specific cases with detail where something is breaking down

OK I'll think this through a bit more carefully when I have a chance. It's a general interop problem which means solutions can be brittle (subject to changing conventions and combinatorial explosion of possible heuristics), and brittle solutions really aren't solutions. For example, I need to find out if "/mnt/x/" where x is the drive letter is common to all WSL distros. If not that could be a show stopper unless there is a sensible rule at play at some level.

If not, then it may be a matter of letting the user provide the logic for translation somehow via configuration (another parameter to deps.tools when reading an edn file for example)

I really want concrete things that don't work and then we can work out from there

conditional code in Windows or even in WSL is something I'm willing to consider but a lot more clarity about the symptoms and examples is needed

OK. I'll at least put the problem up at http://ask.clojure.org to see if anyone has run into it and found a solution. Then you can take it from there.

And I will try other distros, and see if I can also replicate on my Mac/Parallels setup.

I am probably an oddball in working in so many divergent runtimes/os's, but the stuff I'm working on needs to work interchangeably wherever Clojure can run. Clojure's reach is one of the major non-language-specific reasons I love it.

OK, not to try to solve the problem before it's been fully elucidated ... (and now comes the part where I do precisely that)...it might be enough to add a case for :local/root where instead of a string it takes a hash, with possible keys for :os/windows, :os/linux and :os/mac such that it uses whatever the user provides for those cases. It's easy to imagine this solving my problem in any concrete combination. It would eliminate the editing when you need to switch OSs on the local machine, and minimize it as you change your setup. Using local paths is inherently ... local. 🙂 I'll suggest this for consideration in my write up.

sounds ... bad :)

lol, well - like I said - I'll log it and you guys can figure out how to make a good solution 🙂

Auto translating of paths feels - wrought with variables and heuristic pot holes. At least with my suggestion we recognize it's on the dev to point to the right place relative to their local setup on the various "machines" (virtual, emulated, and otherwise). 🙂

But there are probably other ways to skin this particular cat that aren't obvious yet.

when I see all the cats, I will decide :)

👍

I'll keep an eye out as I work on clr.tools.deps, but also will look forward to whatever the Clojure team comes up. (no cats being harmed in the process)