<ocdtrekkie>
Security thought reasonability question. Is it useful to compensate for potential flaws in an IP camera's software by setting an invalid default gateway and DNS address, when you only want to use it on the local network?
<asheesh>
Yes.
neynah has joined #sandstorm
<ocdtrekkie>
That's what I figured. I have an INSTEON camera, which is basically a rebranded Foscam. I'm not sure if my version is vulnerable to the recent HN article issues, but I was thinking it'd be a lot harder to pull data from it if it doesn't know where my router is. And the callouts that were detected sounded like they were by hostname, not IP.
<zarvox>
That is minified, but not gzipped on the wire
<kentonv>
I'd love to get rid of jquery since we hardly use it.
<kentonv>
probably dependencies do, though. :/
<zarvox>
That would be great. The best way to do that would be to drop Blaze, since it uses it in its implementation.
<zarvox>
If we went all-in on React... ;)
<kentonv>
oh, you
<dwrensha>
yeah, where does that jquery dependency come from?
<zarvox>
Meteor uses it for like everything
GeorgeHahn has joined #sandstorm
<dwrensha>
I'm wondering how accounts-identity ends up being 11K.
jemc has quit [Ping timeout: 240 seconds]
frigginglorious has joined #sandstorm
<zarvox>
dwrensha: accounts-identity.html gets compiled into Blaze javascript
<zarvox>
(and somewhat oddly, it appears to preserve whitespace in templates)
<zarvox>
(I guess you have to do that when your templates are strings, rather than structured objects)
<asheesh>
FWIW, the entire bundle presumably has to be loaded to every client before they can see a grain.
<asheesh>
So, like, mrdomino chat is a Rocket.Chat grain, and it's a little sad to load all this stuff before the grain can load itself.
rustyrazorblade has quit [Ping timeout: 264 seconds]
<asheesh>
I don't claim to know yet how we should prioritize decreasing the page load speed, but I wanted to clarify that this affects people who get a grain shared with them and don't really know what Sandstorm is.
<asheesh>
p.s. I do see introjs there, hmm.
rustyrazorblade has joined #sandstorm
axx_ has quit [Ping timeout: 260 seconds]
axx_ has joined #sandstorm
ecloud_ has joined #sandstorm
Jan7 has joined #sandstorm
Try`0xff_ has joined #sandstorm
niekie has quit [Quit: No Ping reply in 180 seconds.]
<zarvox>
does google-perftools require you rebuild your binaries with additional profiling stuff enabled?
<dwrensha>
so then is it so surprising that that's where we're seeing the usage?
<kentonv>
zarvox: you can LD_PRELOAD it
<kentonv>
though I guess spk is statically-linked so that's problematic
<kentonv>
dwrensha: I'm surprised that it is taking significantly more CPU than Ruby
<kentonv>
dwrensha: it may be that there's no way around this because Ruby is just calling stat() too many times in a tight loop, but I think we should look into it
<dwrensha>
I'll continue investigating
<zarvox>
does spk dev cache the output of stat()?
<zarvox>
if not, then we're doing (num_processes)x the amount of work necessary
<zarvox>
(of course, there could be correctness issues if we did it all the time)
<kentonv>
zarvox: no caching is possible. Each stat() is on a unique path.
<zarvox>
even across the three ruby processes?
<zarvox>
they don't ask for the same paths, despite probably loading the same code?
<kentonv>
`spk dev -c` will enable aggressive caching
<dwrensha>
can I do `vagrant-spk dev -c`?
<zarvox>
no, but you can vagrant-spk ssh and run the command that `vagrant-spk dev` prints with an additional -c
<kentonv>
zarvox: I've never actually used flamegraphs. Is the x-axis time?
<zarvox>
width of a bar is how much (relative) time you spent with that frame on the stack
<kentonv>
no, my question is, is the X axis absolute time? Like is this basically an ordered log of what the process did?
<zarvox>
no
<zarvox>
for that you'd need hooks on function enter/leave
<kentonv>
not necessarily
<kentonv>
samples could be arranged in order with the assumption that the shard prefix of the stack is an ongoing call
jemc has joined #sandstorm
<zarvox>
I mean, I guess you could split things up based on offset in the frame (and it's possible that the flamechart *does* organize things in that kind of order)
<zarvox>
but if you make dozens of calls to the same function, then do you want those to be grouped, or split?
<kentonv>
if most time is spent in foo() but it gets called from several different places, how does that show up? Wouldn't there end up being lots of little foo() slivers, which makes it hard to tell that foo() is a hot spot?
<zarvox>
Yes.
<zarvox>
Though other profiling tools tend to be pretty good at showing things like that, but not which callstacks led to them.
<kentonv>
ok. I may be biased since I've used pprof a lot and not ever used flamegraphs, but I feel like pprof output is strictly better.
<kentonv>
pprof output shows call stacks
<zarvox>
more tools more power!
<zarvox>
I should learn to pprof too, sounds like :)
<kentonv>
pprof shows each function as as a box in a graph with edges being calls. Each box shows # of samples where that function was currently being executed and # of samples where the function was on the stack (follow the arrows to find out what was above/below it)
<kentonv>
this is what Jeff and Sanjay threw together when they wanted to optimize. :)
<kentonv>
I used it a lot for optimizing protobufs
<kentonv>
it is open source as part of google-perftools but no one actually advertises how powerful google-perftools is to the outside world. :)
<kentonv>
at Google every binary automatically had the profiler built-in so you could profile any running server in production with one command-line. It was pretty great.
<kentonv>
also heap profiles -- for finding memory leaks, or just where memory is being used. Same visualization turns out to be great for that.
<kentonv>
I definitely agree that profiles without stack traces are worse than useless, fwiw. :)
_whitelogger has quit [Ping timeout: 260 seconds]
_whitelogger_ has joined #sandstorm
XgF has joined #sandstorm
derf- has joined #sandstorm
nwf_ has joined #sandstorm
keturn__ has joined #sandstorm
mrshu has joined #sandstorm
nwf_ is now known as nwf
wolcen has joined #sandstorm
frigginglorious has quit [Quit: frigginglorious]
neynah has joined #sandstorm
KooBaa has joined #sandstorm
KooBaa has quit [Ping timeout: 244 seconds]
ecloud_ is now known as ecloud_wfh
<ocdtrekkie>
asheesh: Lisa16 CFP is definitely already open. In fact, part of it's already closed: Early proposals deadline*: Monday, February 29, 2016, 11:59 pm PST
<ocdtrekkie>
That page is cool, BTW. I probably won't do one this year because I have some specific trips I wanna do this year, and limited vacation time to do them in. But I will keep said list in mind. Might want to start putting 2017 dates on there though, since so many 2016 CFPs are already closed!
_whitelogger_ has quit [Excess Flood]
_whitelogger has joined #sandstorm
<ocdtrekkie>
asheesh: Did I ever show you the awful video from the talk I gave? I forgot if I did or not. I told YouTube to stitch it together.
<ocdtrekkie>
(Finally.)
GeorgeHahn has quit [Read error: Connection reset by peer]
phildini_ is now known as phildini
jadewang has quit [Remote host closed the connection]
mnutt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mnutt has joined #sandstorm
mnutt has quit [Read error: Connection reset by peer]
mnutt__ has joined #sandstorm
jadewang has joined #sandstorm
joshbuddy has quit [Quit: joshbuddy]
_whitelogger has quit [Excess Flood]
_whitelogger has joined #sandstorm
joshbuddy has joined #sandstorm
neynah has joined #sandstorm
jadewang has quit [Remote host closed the connection]
tierce has joined #sandstorm
amyers has quit [Ping timeout: 240 seconds]
tierce has quit [Ping timeout: 276 seconds]
_whitelogger has quit [Ping timeout: 260 seconds]
_whitelogger_ has joined #sandstorm
notevil has quit [Quit: shutdown -h now]
<zarvox>
Hmm. My vagrant-spk Piwik VM, after installing capnproto from git master, can't seem to use the pkg-config files.
<zarvox>
They specify `libdir=${exec_prefix}/lib` but don't specify exec_prefix anywhere.
<zarvox>
Which trips up pkg-config.
<asheesh>
Or 0.5.149 and have the last digit be == the sandstorm release number
<asheesh>
Gotta say, I </3 our lack of capnproto releases.
<asheesh>
0.5.30.149 or whatever.
<asheesh>
Maybe we can make point releases that lack Windows builds, or something, whose release time corresponds to a Sandstorm release time.
_whitelogger_ has quit [Excess Flood]
_whitelogger has joined #sandstorm
_whitelogger has joined #sandstorm
_whitelogger has quit [Excess Flood]
_whitelogger has joined #sandstorm
bodisiw has quit [Quit: This computer has gone to sleep]
tierce has joined #sandstorm
frigginglorious has quit [Quit: frigginglorious]
<zarvox>
dwrensha: yeah, that's the one. Adding the exec_prefix=@exec_prefix@ lines back in fixes thing under autotools. I'm looking into whether that causes a regression with CMake or if it was just an oversight in that commit.
<zarvox>
looks like the latter - the CMake build makes explicit paths of all the variables that autotools makes reference other variables