philip.mcgrath
2018-11-18 17:11:17

Thanks for the very quick fix! I ran into two other things while trying this that might be useful for you to know: 1. On Travis, raco pkg install with its default number of jobs was killed by Travis, apparently for using too much memory. Passing -j 2 worked: I plan to try -j 4 and see if that works, too. 2. I also tried an AppVeyor build, which AppVeyor killed when it had run for 60 minutes, vs. about 10 minutes for the 7.1 and HEAD builds to complete successfully. (That build uses Minimal Racket because I don’t know how to run a Windows installer programmatically, but I could unpack the tarball.)


philip.mcgrath
2018-11-18 17:14:17

Also, I’m guessing "invalid memory reference. Some debugging context lost" is never supposed to happen, but I was sort of surprised that it was an exn:fail:contract specifically.


mflatt
2018-11-18 21:39:45

I expect that RacketCS took too long because it had to compile package dependencies from scratch, and compilation is slow. There may be Windows-specific slowness, too, since the snapshot build for Windows currently likes 6 hours(!). I’m working on that.

Errors at the Chez Scheme level are currently converted to exn:fail:contract by default. There’s more work to do in bridging the two error systems.


carlo
2018-11-19 03:27:26

@carlo has joined the channel


philip.mcgrath
2018-11-19 05:13:43

I can confirm that the Travis build is succeeding now. For comparison, RacketCS took about 25 minutes, vs. 11 minutes for both 7.1 and HEAD—a much more reasonable slowdown—so it definitely does seem like there’s something Windows-specific going on.