Closed Vucius closed 2 years ago
ah,I think I found this reason. there's three rubygems sources:
ffi ~> 1.11
rather than 1.15
so, that's ...BTW, seeing_is_believing's version is 4.0.1
on these three rubygems.
Maybe, something did't go right.
Hi, I'm not sure how to get everything working on all the different systems that need to run it. I'm pretty sure I can get it if I drop Windows support and monkey patch the FFI gem to understand the ARM processors (I do that on my machine so that I can at least use it still). Unfortunately, child process management just doesn't behave consistently across the different operating systems, and the majority of effort involved in this project now is trying to debug that, which is super painful and hasn't yielded much success.
I did rewrite it to just use Ruby's spawn
method, but of course, that failed on Windows, too (🤔 IIRC, it was due to streams behaving differently).
Anyway, I've put bursts of effort into figuring it out a bunch of different times, and kinda feel like I don't know what to do now 😞
🤔 maybe if I made it a little bit dumber (eg not recording which side closed a stream) I could get spawn
to work, which would then presumably remove the need for the childprocess
gem, which could maybe get it back into a functional state.
🤔 oh, maybe also try pointing at the master branch on here instead of at Rubygems. There's some work on master, I think, which I didn't publish b/c it wasn't passing, but it's probably more likely to be correct. Like it should mostly work on Windows, even though the tests don't pass. You might get zombie processes, if you set a timeout and it kill the children, I can't remember if that's still a problem or not, I had it fixed at one point.
Softare:Emacs 27.2
System: Win10
and the error log
Before I installed seeing_is_believing, I installed ffi-1.15.4. But When typed gem install seeing_is_believing, ffi-1.11.3 was still be installed.
I think it is the gem's problem