mnutt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
|jemc| has quit [Ping timeout: 252 seconds]
bb010g has joined #sandstorm
|jemc| has joined #sandstorm
funwhilelost has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mnutt has joined #sandstorm
mnutt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mnutt has joined #sandstorm
|jemc| has quit [Ping timeout: 272 seconds]
neynah has joined #sandstorm
<mnutt>
from the scrollback it looks like at least a couple of people have misinterpreted "you can publish this shared directory to the web" meaning "we'll show a directory listing on your domain"
<mnutt>
unfortunately a) it's not the intended use of the feature, and b) in any case I'm not quite sure how to do it, given that the directory listing is dynamic but web publishing is served out of a static server
<mnutt>
I suppose it would be weird to have a feature "make this shared link a subdomain"
<asheesh>
It's a good feature to have, arguably; it's just that we don't have it.
<asheesh>
You could "just" make your own directory index by crawling the directories and spitting out an index.html.
<asheesh>
But that doesn't help for directories that have an index.html.
<digitalcircuit>
Would that be able to account for files changed via the sync clients?
<asheesh>
Yeah, if the sync code triggers updating those indices. I can't say I love that, but ther eyou go.
<mnutt>
I could have a button "Crawl site and generate index.html" but that would get out of date and get confusing
<asheesh>
You could have a checkbox labled (X) Generate index.html as users sync more files
<mnutt>
I'm not a fan of rewriting the index every file modification, it slows it way down
<asheesh>
which makes the app slower.
<asheesh>
Yeah.
<asheesh>
"obviously" you should "just" be able to connect your Davros file directory to some other app that does want to do that.
<kentonv>
you'd only really need to regenerate it every time a file is created or deleted
<mnutt>
there are some crazier things involving publishing js that embeds a read-only webkey to fetch file lists
<kentonv>
it would be pretty cheap
<kentonv>
accessing a webkey means spinning up the grain and also opens the possibility of security bugs
<kentonv>
I think that regenerating index.html for every file creation or deletion would not be a significant cost
<digitalcircuit>
Perhaps batching generation so it won't run more often than once a minute or so if a lot of files are being added or removed, i.e. initial sync.
<kentonv>
nah, seriously rewriting this file is comparable in cost to what the filesystem is already doing to update the directory
<kentonv>
bandwidth will still be the bottleneck
<kentonv>
maybe if you have a directory with like a million files in it
<digitalcircuit>
Alright
<kentonv>
if you were really worried about it you could have an append-only listing file which index.html fetches using Javascript
<kentonv>
so when syncing lots of files it's an O(1) update for each
<kentonv>
and you could garbage-collect it occasionally based on some heuristic
<kentonv>
but I think regenerating index.html would be fine
<mnutt>
kentonv good points, I will give it a try and see how it goes with larger directories. I also need to do some work where it symlinks all files into /var/www as they're written, rather than symlinking the entire data dir
<mnutt>
btw asheesh not sure if you saw https://www.youtube.com/watch?v=qZlfUJj-Nc0 the other night, one of these days I'm going to get a slick-looking website up on davros.io
mnutt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mnutt has joined #sandstorm
mnutt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
aldeka has quit [Remote host closed the connection]
ocdtrekkie has quit [Remote host closed the connection]
ocdtrekkie has joined #sandstorm
aldeka has joined #sandstorm
uppermgmt has joined #sandstorm
jadewang has joined #sandstorm
mnutt has joined #sandstorm
jadewang has quit [Ping timeout: 272 seconds]
synchrone has joined #sandstorm
frigginglorious has joined #sandstorm
<dwrensha>
ugh, Linode kernel upgrade had me booting in read-only mode. I seem to be back in business after doing s/data=writeback/data=ordered/ in /etc/fstab
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 272 seconds]
mnutt has quit [Quit: mnutt]
gillisigurds has joined #sandstorm
pavar234 has joined #sandstorm
<pavar234>
Hi guys. I am having an issue cross compiling capnproto; it seems it attempts to run a test while doing "make", but obviously fails because the cross-compiled binary will never work there
<pavar234>
./capnp: line 117: /home/build/capnproto/.libs/lt-capnp: cannot execute binary file: Exec format error
<pavar234>
./capnp: line 117: /home/build/capnproto/.libs/lt-capnp: Success
<pavar234>
Makefile:3311: recipe for target 'test_capnpc_middleman' failed
<pavar234>
./capnp: line 117: /home/build/capnproto/.libs/lt-capnp: cannot execute binary file: Exec format error
<pavar234>
./capnp: line 117: /home/build/capnproto/.libs/lt-capnp: Success
<pavar234>
Makefile:3311: recipe for target 'test_capnpc_middleman' failed
<pavar234>
test_capnpc_middleman
<pavar234>
is there a way to prevent the compilation to execute this test?
jadewang has joined #sandstorm
<asheesh>
Hi pavar234
jadewang has quit [Ping timeout: 240 seconds]
<asheesh>
Hmm. I don't know off the top of my head.
<dwrensha>
pavar234: what commands are you using to build capnproto?
<kentonv>
so there's a problem where grains on Oasis get in a bad state where they don't respond to requests even though they are running. The problem fixes itself if you close all open tabs of that grain and wait a couple minutes, which allows the grain to shut down. When it comes back up it is usually OK.
rustyrazorblade has quit [Ping timeout: 276 seconds]
<kentonv>
I'm working on tracking it down
rustyrazorblade has joined #sandstorm
frigginglorious has quit [Quit: frigginglorious]
frigginglorious has joined #sandstorm
notevil has joined #sandstorm
synchrone has joined #sandstorm
decipherstatic has joined #sandstorm
bodisiw has joined #sandstorm
xet7 has quit [Quit: Leaving]
<ocdtrekkie>
I want to use email as a command queue for my home automation system since I'm nowhere near building a web service for it. The theoretical idea is to send mail to the home automation system's email account, and having determined it's from a white listed address, run the command therein.
<asheesh>
That is also sandstormy fwiw!
<ocdtrekkie>
Everything is SSL, and I'm evaluating effectively the From, To, and Subject headers of inbound mail. My home automation system's email address is not known, so that's a point of security, though I assume it's possible to fake that an email came from my address, yes?
<ocdtrekkie>
asheesh: When it has a web component, I want to put it on Sandstorm
<ocdtrekkie>
I was trying to figure out if I could add a secondary authentication without doing anything shockingly nonstandard by putting an authentication key as the display name for my home automation system in my contacts, and then it can evaluate that too.
<ocdtrekkie>
I was wondering if anyone things this is outright insane, or if it's 'good enough'. I don't want to have to authenticate source IPs for different providers and stuff, because that'd be mich more burdensome code.
funwhilelost has joined #sandstorm
<sknebel>
ocdtrekkie: shouldn't a properly configured mailserver do that by itself nowadays, checking SPF and stuff?
<sknebel>
also, you could always require all mails to be GPG-signed, if you have that set up on all devices that might send commands
<ocdtrekkie>
My home automation client isn't a properly configured mail server, so I don't want to assume.
<ocdtrekkie>
It's POP3 implemented in a couple dozen lines of Visual Basic, and thank God I found a code sample ro do it over SSL.
funwhilelost has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<ocdtrekkie>
And I have exactly zero devices set up to GPG sign my mail.
<kentonv>
short of PGP, all methods of authenticating email are sketchy.
<kentonv>
if your recipient address is unguessable, and you only email it from trustworthy sending servers, and both ends support TLS, then you probably don't need to worry about authenticating received emails
funwhilelost has joined #sandstorm
frigginglorious has quit [Quit: frigginglorious]
frigginglorious has joined #sandstorm
<ocdtrekkie>
kentonv: email address isn't posted, but it wouldn't meet 'unguessable' standard. I was figuring a key in the To: Display name would be though, and its easier to rotate as well.
<kentonv>
ocdtrekkie: Yeah I suppose that would work
<kentonv>
You could also put the key in the body, say on the first line, or something like that. Might be less weird.
xet7 has joined #sandstorm
<ocdtrekkie>
kentonv : Then i have to type or paste it, or rely on clients with a template of some sort. This could save in Contacts.
<kentonv>
fair enough
<ocdtrekkie>
Also, I save a lot of processing by readline-ing the four header lines I want and then dumping the rest outright. My code never looks at the message body.
* kentonv
goes digging into the Blackrock code to fix an issue, discovers a TODO comment he wrote a few months ago describing exactly the problem.
<dwrensha>
TODO(wrong)
notevil has quit [Quit: updating ...]
xet7 has quit [Quit: Leaving]
xet7 has joined #sandstorm
notevil has joined #sandstorm
nicolagreco has quit [Remote host closed the connection]
TheJ has joined #sandstorm
<TheJ>
ping
<kentonv>
hi
<TheJ>
I've been trying to get sandstorm running on a fresh ubuntu 15.10 box. having troubles geting it to use anything other than local.sandstorm.io
<TheJ>
I've been digging throught eh docs extensivelyu...
<kentonv>
what problem are you seeing?
<TheJ>
when I change teh config file's base-url to anything other than local.sandstorm.io I can't even reach it on the local box. not even an ip will work.
<TheJ>
I want to use my own domain ( or a sandstorm.io domain. whichever)
<kentonv>
hmm, what is BIND_IP set to?
<TheJ>
it was set to 127.0.0.1 until I foudn teh Troubleshooting FAQ, then changed it to 0.0.0.0
<kentonv>
ok, 0.0.0.0 is correct
<kentonv>
when you try to reach it by IP, does it fail to connect, or does it return an error page?
<TheJ>
unfortunately I'm not at the box now....
<TheJ>
IIRC it would fail with an IP but give a page if I used a domain name
frigginglorious has quit [Quit: frigginglorious]
<TheJ>
What are the critical lines in teh config to set it up with a domain? I see "base", "Wildcard", "Bind" in the docs...
<TheJ>
the generated admin-token page also would not load.
<kentonv>
they are all important... BASE_URL sets the main address, WILDCARD_HOST identifies a DNS wildcard that is also mapped to the server (sandstorm needs this to generate new random hosts for security reasons), and BIND_IP specifies which network interface to listen on (0.0.0.0 means "all")
<kentonv>
also PORT is important, of course
frigginglorious has joined #sandstorm
<kentonv>
if you've set BIND_IP to 0.0.0.0 then you should be able to connect via http://<ip-address>:<port>
<kentonv>
(after "sudo sandstorm restart")
<kentonv>
you'll get an error page that starts with "Error looking up DNS TXT records"
<kentonv>
if you don't see that then it seems like something is wrong at the network level, but I'm not sure where to start debugging that...
<TheJ>
I will try again tonight. Yes, i was restarting the server. I did get that eror page
<TheJ>
I quite certain it is my config. This is a completely fresh install of EVERYTHING (even a new HD)
<kentonv>
oh OK, if you got to that error page then the network is good. Now you need to make sure that BASE_URL matches exactly the URL in your browser
<kentonv>
(except you can't use an IP address)
<asheesh>
xip.io!
<TheJ>
ok. so it must use a domain name in teh base and wildcard lines, correct?
<kentonv>
yes
<asheesh>
(or sandcats.io)
<kentonv>
and the wildcard line needs to have a * in it, and that needs to be configured as wildcard DNS, meaning the * can be replaced by anything and it still maps to your server
<kentonv>
most DNS providers support this, but it's not normally configured by default
<TheJ>
..and so long as the eth port has a DNS server listed it should find the server
<TheJ>
yup. I kept the * in that line
<kentonv>
no, the eth port's DNS is for outgoing lookups. You need to configure a DNS server somewhere to answer lookups for your domain -- this won't happen automatically, unless you use sandcats.
<kentonv>
e.g. most domain registrars have a DNS service you can use, but you'll need to configure it
<kentonv>
if you'd rather use sandcats and automate it, it's probably easiest to reinstall from scratch. Enabling sandcats after-the-fact is a little tricky at the moment.
<TheJ>
is there an info page for sandcats? The "Free sandcats.io HTTPS" docs don't actually say where/how to get one
<asheesh>
Oh! "how to get one" indeed.
<TheJ>
oh ok
<asheesh>
The way you get one is as part of the install script.
<asheesh>
I can add a doc soonish on getting one outside of the install script, but admittedly the install script is the only supported way at the moment.
<TheJ>
ahh...I must have glossed over it as I was wanting to use my own domain. I really don't care. I just want it online ;-)
<kentonv>
if you don't have any important data in Sandstorm currently, then you can nuke your existing install with "sudo sandstorm stop" and then "rm -rf /opt/sandstorm".
<TheJ>
cool! That's what I will do and start again ;-)
<kentonv>
ok, let us know if there are any more problems!
<asheesh>
Awesome TheJ. And let us know if it does work!
<TheJ>
PBKAC - Problem between keyboard and chair ;-)
<asheesh>
Nah - we could have improved the doc there. I'll work on adding that now.
<TheJ>
as far as the docs... an example config file would be awesome !
<asheesh>
Oh! Interesting. Yeah! OK.
<kentonv>
ideally we want to make it so people never have to edit the config... but we're not there yet.
frigginglorious has quit [Quit: frigginglorious]
<TheJ>
also, somewhere i found "sudo services sandstorm restart" which doesn't seam to restart the server.
<TheJ>
"sudo sandstorm restart" found elsewhere in teh docs certainly does
<asheesh>
sudo service sandstorm restart # singular service - does that work?
<asheesh>
Interesting. What happens when you do the 'sudo service sandstorm restart' command?
<TheJ>
I would have to try tonight. I'm not at my server right now
ocdtrekkie has quit [Remote host closed the connection]
<TheJ>
nothing. the shorter of the two states that the server is restarting
aldeka has quit [Remote host closed the connection]
<kentonv>
it should work if you're using sysv init, but these days a lot of distros have switched to systemd, which is probably different
aldeka has joined #sandstorm
ocdtrekkie has joined #sandstorm
<TheJ>
I have no idea. It's a fresh Ubuntu 15.10 install
<kentonv>
well, "sudo sandstorm restart" should work regardless of init system
<TheJ>
it did :-)
<asheesh>
Except it might mess up the service state!!
<asheesh>
But OK.
<asheesh>
Hmm, I guess it won't.
<asheesh>
I should try that, brb.
<asheesh>
Yeah, systemd + 'sudo sandstorm restart' seems fine. I don't know why I thought it wouldn't be fine.
<asheesh>
I guess I should go change the docs to all say 'sudo sandstorm restart'!
<kentonv>
"sudo sandstorm restart" actually sends a restart request to the Sandstorm server, and it restarts the subprocesses, but the root process doesn't restart, so systemd wouldn't even notice anything happen.d
<TheJ>
further on teh docs... an example config file for the HTTPS setup would be awesome as well
<asheesh>
TheJ: Your feedback is pure gold! Thank you so much! This is great.
<TheJ>
I've written TONS of documentation !
<asheesh>
I must AFK for 30 min for a meeting, but I'm glad you are here TheJ and any other feedback you have, I'd love to hear it. I'm the main docs maintainer, I suppose, so you're definitely talking to the right person.
<TheJ>
I've found showing teh big picture first then describing the details works very well.
<TheJ>
Cool. I have to go to but I will touch base again will any refinments I can think of ;-)
<asheesh>
Cool.
<asheesh>
Most importantly, let us know if you get your server fixed.
<zarvox>
reactive template rerendering doesn't take the tree shape into account
<dwrensha>
tree shape?
<zarvox>
basically, it'll rerender children when anything about them changes, even if the children would (in a pure, from-scratch render) be removed from the tree entirely
<zarvox>
the upshot of this is that when I add an onDismiss to the powerbox request popup to clear the current powerbox request, there is a render where the popup is still being rendered but the request info is now undefined
<zarvox>
which produces a bunch of exceptions on the JS console
<zarvox>
possibly this has to do with our use of Template.dynamic in the magical topbar stuff
decipherstatic has quit [Remote host closed the connection]
<dwrensha>
I would not be surprised!
<zarvox>
but nonetheless, this is a thing that just doesn't come up in React-land, because the component lifecycle is so explicit
<zarvox>
so I guess the question is "can you safely nest reactive contexts?" and "if so, when you nest them like that, which context should get rerun first?"
nicolagreco has joined #sandstorm
<zarvox>
and "can you even express that to Tracker"
notevil has quit [Quit: ...]
<zarvox>
I guess I can maybe work around this by explicitly stopping that reactive recomputation in topbar when onDismiss() is called
nicolagreco has quit [Remote host closed the connection]
bodisjw has joined #sandstorm
bodisiw has quit [Ping timeout: 240 seconds]
<simonv3>
Hey all, I’m sending someone who’s not a Sandstorm user an Etherpad grain, and it’s telling them that it’s a demo and that it will expire in an hour.
<asheesh>
simonv3: ... it's because they're logged in using a demo account.
<asheesh>
I believe.
<asheesh>
That's a fascinating edge case though.
bodisjw has quit [Quit: Leaving]
<asheesh>
If they did:
<simonv3>
Yeah, I linked them to the app earlier to let them try, so that’s probably why
<asheesh>
... right
<simonv3>
(they probably tapped through on Demo)
<simonv3>
and then I was like “wait, it’ll just be easier if I create the grain and share it with them"
<asheesh>
Honestly the point of the countdown is to motivate people to make an account.
<asheesh>
The secondary point of the countdown is so that if people really don't want to sign in, they know that their data really will go away.
<simonv3>
Nothing’s going to happen to the grain cause it’s mine though, right?
<asheesh>
In this case, the secondary point is meaningless since it's a shared grain from you.
<asheesh>
Exactly.
<simonv3>
cool
xet7 has quit [Ping timeout: 240 seconds]
<zarvox>
None of the things I tried in topbar.js seem to be fixing this, so I guess I'm going to make all the powerbox template helpers deal with an undefined data context
xet7 has joined #sandstorm
<zarvox>
I wish I knew why Tracker was flushing eagerly and why stopping these reactive computations isn't enough to keep stuff from rerendering when it really doesn't need to