webrender is pretty smooth for me browsing around right now. A+, would void my warranty again.

hlieberman: neat!

I'm using bindgen in a fairly large project, and I've noticed that it generated accessor methods for the bitfields on some structs, but not for bitfields on other structs

db48x: do you have a test-case by any chance?

yes, but not a very small one

ok, I've got a fairly small one

emilio: my creduce isn't working, but gist.github.com/db48x/ac5185816f1b921bb8ca804883165e17 is a self-contained example

db48x: that's reasonably sized to debug, mind filing an issue? I'll try to get to it

and of course I can't compile creduce either :P

test result: FAILED. 259 passed; 160 failed; I'd say this synup is going well ._.

Thread 'header_whitelist_namespaces_hpp' panicked at '"::" is not a valid Ident'

This is going to be so much fun

hey, creduce is doing things

The file being reduced has reached zero size; our work here is done.

standups: birthday PTO

ferjm: Happy birthday !

ferjm: 🎂

eijebong, est31 thank you! :)

judging from their social network nick, it might also be nrc's birthday ? :)

emilio: Welp, can't update quote without having those ugly #[doc=] :'(

emilio: paste.bananium.fr/?6177a4d8bb0a44b7…7dAg0jhRkU3EvIEzAVJLDZZjzQ6OrHyEsY=

I used TokenStream::from_str("/// the comment") hoping that it would keep it this way but it didn't

standups: catching up from pto+bank holiday

ferjm: Happy birthday!

eijebong: :(

eijebong: I don't really think that's acceptable :/

emilio: Unfortunately that means that we can never update quote/proc_macro2 ever again...

And I don't think that's acceptable either :/

emilio: I might have an idea, I'll try it

Well, it compiles, let's see if it does what I think it should do

emilio: I can emit ///, now I have to find out how to insert a \n, which I'm not sure I can do

eijebong: we should probably just ask the syn people for an API to do this

eijebong: seems better than hacking around it

dtolnay: ^

emilio: I feel like this is at least the third time we ask for something like that :(

emilio: Welp, we can't even emit that anyway. There's no way to add a \n nor a string literal without quotes

emilio: We could add another pass that changes #[doc="pouet"] to /// pouet but it's meeehhh

eijebong: yeah, that'd be kinda lame :(

emilio: I'm afraid we won't have another option :/

cybai: thanks!

standups: submitted rust-lang/rust #51165, will hopefully unblock rustups

ferjm: oh, happy birthday as well :-)

SimonSapin: Do you think that the syn segfault might be related ?

eijebong: no idea, is it in thread-local storage code?

SimonSapin: Do I look like I know ? :p

It was the main thread overflowing its stack for no apparent reason

even with a 128 MB stack?

SimonSapin: Yeah, we set the cssparser stack to 128MB to try and it still appeared

thank you, emilio :)

hey, it looks like it might be doing useful things

crowbot: infrastructure report

standups: rust impl days, day 2. More GlobalAlloc rust-lang/rust #49668, main remaining blocker is what to do with zero-size allocation requests

standups: performed outstanding gecko reviews

standups: performed outstanding gecko reviews

standups: Getting mozjs_sys to cross-compile for android.

Yay, got a cross-compile working locally. The fixes were servo/saltfs #844 and servo/rust-mozjs #427

Hopefully these are the last blockers for getting bindgenup to work in CI.

which is the other blocker

jdm: having non-borked build machines helps too :)

jdm: do you know why we're getting errors at travis-ci.org/servo/saltfs/builds/385259720? Looks like networking issues :/

ick

yeah, just intermittent network stuff

pressing the retry button usually helps

restart button, that is

jdm: yay, pushing random buttons till things work!

standups: reviewed FileReaderSync PR and unhandled promise rejection PR

standups: Landed servo/saltfs #844 and servo/rust-mozjs #427, which are hopefully enough to get bindgenup to cross-compile for android.

friiiiiiiiiiiiiig

the new servo-linux3 is broken in the exact same way the old one was

before I "fixed" it

arrrrrrrrgh

jdm: lol

CI was a mistake

so, if pip segfaults

and upgrading pip uninstalls salt

what options am i left with?

uninstall pip

gah

 cat isolated-test-case.h

struct glyph_row;

emilio: can you suggest a way to ensure that this thing has at least one stupid bitfield?

haha

jdm: is servo-linux-cross1 still on the old saltfs? I'm still getting cross-compile errors build.servo.org/builders/android/bu…ilds/11630/steps/compile/logs/stdio

ajeffrey: yeah, I haven't deployed the new change

ajeffrey: hence the S-needs-deploy

jdm: OK, let me know once it's live

I guess --bindings-grep 'pub _bitfield_1' might work

standups: Fixing bugs in my queue (gaol, webrender, pathfinder).

ajeffrey: deployed

standups: Fixing bugs in my queue (gaol, webrender, pathfinder).

jdm: ta

wat

why is servo-linux3's current build not failing like the previous four builds

what is happening

good question

Oh pox build.servo.org/builders/android/bu…ilds/11631/steps/compile/logs/stdio

looks like something didn't get the right sysroot

pcwalton: would you like to review servo/ipc-channel #199 ?

jdm: looking

jdm: I couldn't reproduce that Travis CI failure :(

even on that version of Ubuntu with stable rust

is it intermittent?

pcwalton: it's consistent on every PR that gets opened

I'm confused then

jdm: I guess I could add some logging to see why unshare() is failing?

the downside of having code that only one person on the team knows anything about

well, the problem is that unshare is returning some kind of error code

knowing what the error code is would help :)

jdm: r? servo/gaol #48

standups: wrote fix for exceeding ipc-channel message size limit

pcwalton: looks like it spits out -1

pcwalton: so, I guess we need to look at errno somehow

oh right, it's errno

bah

jdm: pushed an updated version

thread 'main' panicked at 'unshare() failed: Os { code: 1, kind: PermissionDenied, message: "Operation not permitted" }', platform/linux/namespace.rs:334:9

Hmm, that last build failure was caused by bindgen not passing CXXFLAGS to clang,

it's fixed by asajeffrey/mozjs d1901a3

emilio: Do you know why bindgen doesn't use CXXFLAGS when building from a .hpp file?

standups: investigated new webgl regression

db48x`: If you're looking for creduce tips, fitzgen is your person. I suspect we may just think we can't represent it because of alignment or something and bail out and make the whole thing opaque...

ajeffrey: I guess nobody wrote the patch. In general we try to mess with the environment as little as possible...

emilio: fair enough, I just passed them in by hand.

jdm: is there any test-case for #20877 or something like that?

jdm: Looks good to me in any case

Threads are hard :(

emilio: not without adding a test-only API that panics the webgl rendering thread, and I'm not even sure if that would yield testable behaviour

jdm: oh well, fair enough. r=me then

thanks!

Hmm, a segv running script codegen on a mac: build.servo.org/builders/mac-rel-wp…uilds/8674/steps/compile/logs/stdio

ajeffrey: yeah, that's the ongoing #20796

the best part is how macstadium reboots it a couple times a day

and then it works for an indeterminate number of builds

and then it starts falling over

jdm: ta

OMG!!! build.servo.org/builders/android/builds/11632

bindgenup built on android! My cross-compile woes may be coming to an end!

woo!

ajeffrey: should we file an issue on bindgen for servo/mozjs #136 ?

jdm: indeed

jdm: well, I'm not sure about this one -- it is a bit of a gotcha, but I can also imagine cases where it's the Wrong Thing To Do.

standups: Got bindgenup to cross-compile to android. build.servo.org/builders/android/builds/11632

gw: any idea why passing GL_MAX_VERTEX_UNIFORM_VECTORS to glGetIntegerv would yield INVALID_ENUM?

standups: set up my vr environment at home

jdm: checking

jdm: this is on desktop, I presume?

gw: yeah

jdm: It doesn't appear as a valid param to glGet in docs.gl/gl3/glGet (desktop gl3), but it does in docs.gl/es3/glGet (mobile/es3) - just checking why now

eh?

huh, you're right

this is awkward, since it's defined as a valid value for webgl

jdm: looks like on desktop GL, you probably need to use GL_MAX_VERTEX_UNIFORM_COMPONENTS (which is scalar components rather than vectors)

oof, the GetParameter implementation is complex

jdm: although, it's not necessarily a simple conversion between those two, due to packing rules (although in 99% of practical cases the simple 1:4 ratio would work)

aieee searchfox.org/mozilla-central/sourc…as/WebGLContextValidate.cpp#523-558

jdm: ah, so gecko just assumes 1:4 ratio, cool

jdm: I have a suspicion about #48 -- commented (tl;dr I wonder if Travis is itself using namespaces to run its jobs, which is interfering with our attempt to create new ones)

hmm

it's logical that Travis would be attempting to sandbox its tests

in case they rm -rf / or whatever

I have literally no idea what the context is, but I thought the default travis images used docker these days c.f. blog.travis-ci.com/2014-12-17-faste…with-container-based-infrastructure

does anyone here know of a project that uses bindgen and builds on osx via travis ci?

db48x`: the mozjs_sys crate does, as of last week :)

ajeffrey: aha

db48x`: osx was pretty painless, the problem cases were Windows and cross-compiling for Android.

yes, your .travis.yml file is practically boring

ours is awful

hmm. in your appveyor config file you set a giant path up front

I like that idea

standups: will finish writing nav timing tests in the morning