jeapostrophe / pict3d

3-dimensional picts
GNU Lesser General Public License v3.0
38 stars 9 forks source link

Documentation link broken #11

Open Linchy opened 7 years ago

Linchy commented 7 years ago

The link on the front page (https://github.com/jeapostrophe/pict3d) is broken:

http://pkg-build.racket-lang.org/doc/pict3d/index.html

Page not found

((uncaught-exception-handler) ((+()((+()()()()())(+()()()()())(+()()()())))(+()()()(*)))) uncaught exception: 404

jeapostrophe commented 7 years ago

Hi Lincy,

The doc building service sets a limit on memory usage and pict3d exceeds it while compiling: http://pkg-build.racket-lang.org/server/built/fail/pict3d.txt

@samth Is there any hope of improving TR's performance on pict3d?

@mflatt Any hope to get a higher limit? I'm not sure what it would need. Maybe a way to flag in the set up some special exceptions for useful packages?

samth commented 7 years ago

I think it's unlikely that all of pict3d and its dependencies can be made enough faster to fix this on Typed Racket's side.

mflatt commented 7 years ago

@samth Have we figured out why the pict3d build got dramatically slower in the last year or so?

samth commented 7 years ago

I think I figured it out, but I no longer remember. Let me see if I can dig that up.

Linchy commented 7 years ago

@jeapostrophe if it's any help then, I had to increase my DrRacket memory limit to 512mb before any pict3d samples would compile.

samth commented 7 years ago

@mflatt Looking back at our investigation, it appears that we didn't figure out why this happened and it didn't reproduce outside of the pkg-build environment.

samth commented 7 years ago

Also, the documentation successfully builds on HEAD here: https://plt.eecs.northwestern.edu/pkg-build/doc/pict3d@pict3d/index.html

mflatt commented 7 years ago

@samth some things are starting to come back to me... I think we discovered (as @Linchy hints) that the issue wasn't so much speed by itself as memory use. Robby added more memory to his virtual machines, and I did the same temporarily to build once; in its default configuration, the build VM thrashes.

Were we able to replicate the difference in memory use outside the pkg-build environment?

samth commented 7 years ago

My email suggests that a -j 1 build of pict3d maxed out at 1.5GB on a 6.8 pre-release test.

samth commented 7 years ago

I can re-do tests for that if it would be useful.

mflatt commented 7 years ago

I don't expect the test results to change, but 1.5GB to compile a library sounds like something is wrong. If I remember correctly, v6.7 is the same, while v6.6 used much less memory, so maybe we can track down the relevant change.

mflatt commented 7 years ago

@samth When I try the current development version or v6.5 with env PLTSTDERR="debug@GC error" raco setup -j 1 pict3d, I get a peak memory use of about 1GB for both. (The development version is slightly lower.)

Before that command, I started with pict3d installed but deleted its ".zo" files. I get a similar peak when starting with a fresh snapshot and use raco pkg install -j 1 pict3d.

So, unless you're seeing something different, I guess peak memory use hasn't changed recently, after all. I'll make sure the VMs have more memory, and probably that will solve the problem.

samth commented 7 years ago

I haven't experimented recently, but I would also expect the current development version to be faster than 6.5 in general, so your data is consistent with that.