Update libsass to fix materializecss crash

[tony]

I know you just tagged a release. There is a fix effecting people with https://github.com/Dogfalo/materialize that only got merged in within the last 24 hours.

This fixes a segfault that happens while compiling materializecss.
Its tested with https://github.com/torchbox/django-libsass

The bug is documented on the libsass website at
sass/libsass#1960

The patch was merge upstream at
sass/libsass#2032.

[coveralls]

Coverage remained the same at 94.464% when pulling f45b3c2 on tony:update-libsass into a3748d8 on dahlia:master.

[asottile]

We don't do unstable releases any more, but I will release a version targeting 3.3.6 when I get home :)

[tony]

I was concerned about that, its nice 3.3.6 covers that PR. I amended the PR and tested it at the 3.3.6 tag.

[coveralls]

Coverage remained the same at 94.464% when pulling 24a86f0 on tony:update-libsass into a3748d8 on dahlia:master.

[asottile]

The appveyor failure is strange... I wonder if they broke something upstream. (I'm on my phone or I'd debug it)

[tony]

%PYTHON%\python.exe -m pytest sasstests.py
============================= test session starts ============================= 
platform win32 -- Python 3.5.1, pytest-2.9.1, py-1.4.31, pluggy-0.3.1 
rootdir: C:\projects\libsass-python, inifile:  
collected 89 items 

sasstests.py ...........Command exited with code -1073740791

I can't re-run the build.

[tony]

@asottile can you rerun the build and see if its not a fluke?

[asottile]

Weird, I don't have permission for some reason -- you can retrigger a build with a trivial amend:

git commit --amend && git push origin HEAD -f

I have a feeling this is a real failure -- I've started debugging it locally

[tony]

Good idea, for me it'd be git push tony HEAD -f. Lets see what it says

[coveralls]

Coverage remained the same at 94.464% when pulling fa7aeda on tony:update-libsass into a3748d8 on dahlia:master.

[asottile]

It crashes for me locally as well -- but only for 64bit python3.5 -- I'll dig into this further

[tony]

Ok, this PR is just an update, if you have something, feel free to close this.

[asottile]

I'm trying to bisect when things go wrong. From some googling it appears to be a stack overrun and it only seems to happen in error cases (and curiously only on python3.5-64 -- maybe that means it's our code's fault somehow?).

[asottile]

This seems to be the first commit which triggers this strange behaviour: sass/libsass@527f3a8

[asottile]

Created sass/libsass#2046 with a reproduction which doesn't involve python

[asottile]

@tony looks like we figured it out :) sass/libsass#2046 sass/libsass#2048

I think we'll be skipping 3.3.6 and going straight to 3.3.7 when that comes out

Thanks for the PR though, sorry it didn't work out due to other reasons :)

[tony]

Ok awesome.

I see what was happening there

[asottile]

upstream 3.3.6 is now available as 0.11.2! ✨