This is what I’m confused about
Yesterday, I git pull; make and it failed due to the htdp error.
Today, I did it again, and despite the fact that there’s no bug fix, somehow the error is gone.
See the 55733 build vs 55734
I have no idea what’s going on
That’s strange indeed.
I usually build a racketcs per day and I couldn’t build it due to the error.
let me try again.
It might depend on how you install it. I have found that sometimes I run raco pkg install ... and it fails during raco setup. But if you issue raco pkg install .. again, it says the package is already instealled and raco command succeeds.
This has driven me insane in the past.
So, if you have installed racket and went for htdp user scope, then the same happened.
I always use installation scope to try to mitigate these things.
Also, I just noticed that https://snapshot.racket-lang.org/ci-snapshots/ has stopped working since June
oh wow
hadn’t noticed that.
are those generated by ci?
the snapshot job works on a regular basis, so unsure what’s going on here.
Another possibility is that the snapshot job works fine, but the webpage is not updated properly
The only thing the job does is an s3 sync: https://github.com/racket/racket/runs/1317778579?check_suite_focus=true#step:7:1
which succeeded.
I am honestly unsure how it work.s
Surely @samth will know straight away
@sorawee by the way, my CS daily build failed again.
so problem is still there.
Unsure what the policy is here but I think we need to be clear that these breakages shouldn’t happen.
@samth @mflatt can we revert the breaking commit to htdp until Mike has a chance to fix it?
@pocmatos is it the make-result is unbound error?
I just found that the problem was fixed at https://github.com/racket/deinprogramm/commit/93b066ef877afe3440045b71e607de81ee7e37da
could it be the case that your deinprogramm is not up-to-date somehow?
Yeah, and this explains the weirdness that I saw.
The commit in racket/htdp was merged, but it depends on a change in deinprogramm which had not been merged yet.
Then, I git pull; make between these points, so I got the errors.
But now that the change in deinprogramm is merged, git pull; make works fine once again.
yes - it’s the make-result unbound error
deinprogramm?
oh my brain translated the german and I got confused “could it be the case that your "your program” is not up-to-date somehow?" lol sounded funny.
lololol
I don’t even know what deinprogramm means
but I don’t understand, racket should be downloading deinprogramm from master.
that’s true
so, it should be getting the most recent version.
when did the fix went in?
1am yesterday
(PST)
hummm, I am confuse.d
I will need to look again.
I am quite busy today so probably won’t have time, lets see.
So things that could be improved is that these merges should have happened at the same time.
Matthias merged a change in htdp at 4pm, Oct 27.
right.
but it needs a change in deinprogramm which was merged later at 1am Oct 28
Another thing that got me thinking is… I never used htdp or deinprogramm . I am compiling the very base of racket. Why is racket even trying to compile those…
Oh - I figured it out…
Do you build DrRacket?
DrRacket requires htdp, I think
Nope - no DrRacket, but I know what it is…
make install goes to your $HOME/.racket/7.9.0.3 and runs raco setup on all of it.
So, I kept getting the error on the broken packages I got installed.
I did an rm -rf $HOME/.racket/7.9.0.3 and now racket compiled and installed fine.
grrrr….
That’s a bit annoying.
ah
Mainly because I thought I am always using installation scope, so .racket shouldn’t even be populated.
I should look into this but it’s one of those things that you postpone cause it’s not too bad…
@pocmatos: it looks like xvfb-run is not available in https://github.com/racket/htdp/pull/114
which is weird. I copied it from https://github.com/racket/drracket/blob/master/.github/workflows/ci.yml and it seems to work fine there
One difference is ubuntu version. Another is that one is in a container and the other is not.
Do you know how to fix the problem?
Sorry @sorawee didn’t know xvfb-run was necessary.
So xvfb-run is not installed in the container. We should update the container to have it.
I am now rebuilding the container and will push the new image soon after testing. Give me a couple of hours. Your PR should just transparently work.
@sorawee it should now work. I don’t have permissions to re-run tests on htdp. However you can force it by force-pushing your commit.