do {blog();} while (j == awesome); » BigBlueButton

FF you, FFMPEG

Justin — Fri, 25 May 2012 15:52:57 +0000

So we’ve decided to move from using FFMPEG ourselves over to using Xuggler, which in turn uses FFMPEG as part of it’s internal processes. We’re so close to getting it to work that we can just about taste it; we know where to find the RTMP URL coming out of BigBlueButton and confirmed with FlowPlayer that the stream is working. The thing is, the same RTMP URL won’t work in Xuggler; the IContainer.open method returns an invalid entity exception.

Chad’s theory is that we need to add more metadata to the RTMP stream before it leaves BBB, so that Xuggler can read it. The reason FlowPlayer could read it without problems was because it’s expecting a Flash video to begin with. Now the question is, what kind of metadata do we need to add in order to get Xuggler to accept it?

FFMPEG: A “very active project” that nobody ever talks about

Justin — Wed, 23 May 2012 18:04:24 +0000

Show of hands: Who here has ever heard of FFMPEG?

Apparently, it’s our best bet for converting live Flash video into non-Flash, HTML5-compatible video on the fly. And yet, all the information Chad and I find about it is either years out of date, about converting video that is being streamed from a file, or otherwise completely irrelevant to our needs. But the one consistent theme is that FFPMEG is “a very active project, lot’s of changes and releases all the time.“

There is one tutorial in existence for using FFMPEG, where it is being used to do something other than what we want. The documentation is well-hidden and unhelpful. Everyone insists that it can be used for live, on-the-fly encoding, but nobody is saying how. The few people who’ve said that they HAVE made something like this, regardless of whether it’s with FFMPEG or not, can’t tell anyone because the rest of the project was proprietary. Stack Overflow information on the subject dies in late 2010 or early 2011. Are we actually going to have to write our own video encoder/decoder from scratch?

Moving forward

Justin — Tue, 15 May 2012 17:14:48 +0000

We finally have a meeting to decide what we’ll be working on this term! Well, Fardad has the meeting. Chad and I are just here reading up on WebRTC in case the project turns out to be “convert BigBlueButton to HTML5″. And let me tell you: I am lost. I’m the kind of person who needs to actually play with the code and see what THAT lever does before the documentation really starts to have any meaning for me, so at the moment it’s a bit daunting. But it’ll be fine, I’m sure. I started my first work term staring at the BBB documentation exactly the same way, and look at how that went!

Going off the script

Justin — Thu, 10 May 2012 17:08:20 +0000

So, yesterday’s scripting experiment didn’t exactly work out. I was trying to get the Dev Install script to automatically handle it when, during the build process for the locales or the client, Java encountered a runtime error. This is an intermittent problem when running those builds, really the only way to handle it is to run the command again. Currently, the way the script handles it is to ask the user after each build whether they see a BUILD SUCCESSFUL message or not (conveniently located immediately above the question)

My thinking was to redirect the output and error streams of the build commands to a log file, and grep that log file for the runtime error message. And that part worked really well! Except….

See, there’s another intermittent error that I’d forgotten about, which I just call “the build.xml hiccup.” Basically, sometimes build.xml just plain doesn’t like what it sees on line 318. Just like the runtime error, if you run build a second time without changing anything, it works properly. The dilemma arises from a few facts:

A) The error text for the “hiccup” is the same as the error text for a REAL build.xml error, where something is configured wrong or simply shouldn’t be there.
B) As it is an intermittent error, it’s conceivable that it could happen twice in a row.
C) Moving forward, there is no guarantee that the “hiccup line” will stay at 318, so I can’t hardcode an “ignore line 318 clause”
C2) Also, what if someone makes a legitimate error on the “hiccup line” when developing build.xml?

In short, while I can easily isolate the intermittent runtime error, I can’t get the script to distinguish between legitimate build errors and “hiccups.” Automating the handling of one while leaving the others unaddressed just feels slipshod to me, and the prompt that is already in the script already covers all eventualities with a nice, broad “Did it work?” sort of question.

That isn’t to say the script is perfect. If you do have a legitimate build error, no matter how many times you say “No, it didn’t work” the script can’t fix it and will continue to loop. However, remember that this is the script for installing the module onto a development environment. Someone who is developing the code should recognize the build error as something to be looked into after seeing the same one several times, and know to Ctrl-C out of the script. Like I said, not perfect.

Anyway, that’s my explanation of yesterday’s little experiment.

I wonder if there’s room for Klingon…

Justin — Wed, 09 May 2012 14:23:36 +0000

So! Not only is the BigBlueButton polling module ready for public consumption, with an easily accessible repo and installation scripts, but within hours of posting it to the BigBlueButton dev forum with a call for people to try it out and help translate the locale files we had eight translations ready to go! The power of open-source, people.

If you’d like to have a look, the repo is here (just follow the directions in README) and the localization website is here.

On a slightly different topic, I had a thought just now. Currently, the installation script for a development environment prompts the user after ant locales has run, and again after ant, because of the possibility of a Java Runtime Error inturrupting the process; the prompt reads “Was the build successful?” and if the answer is No, it repeats whichever command it just attempted. I’m considering updating the script to redirect all output from these commands into a temporary file, and basing the repeat condition on whether a grep finds “BUILD FAILED” or not in that file.

The reason I’m hesitant to do it is, sometimes there is a very good reason for the builds to fail; it’s not always just a randomly-occurring Java error. Although I could have the grep look for a piece of text from that Java Error….. Any thoughts?

Must not curse at the repo. Must not curse at the repo.

Justin — Fri, 04 May 2012 18:20:15 +0000

I think we may have conquered GIT. Since Anatoly left for Red Hat, we decided to use his old repo as the master and have Chad and I create our own forks off of it, mostly as practise so we don’t run into a million conflicts like we did with the LDAP work.

I guess I should mention, we’re revamping the myprogrammingblog repo to be an “installation” repo for the Polling Module.

Anyway, here are the steps. Mostly for our own reference later.

1) Go to the GitHub page and click FORK.

2) In a terminal, type git clone git@github.com:/

.git

3) cd

4) git remote add upstream git@github.com:

.git

5) git fetch upstream

This gives you a fork from the main project (you can click “My Fork” where the FORK button used to be to see it), you can commit to that with the usual add-commit-push system. When you want to push your fork to the main project, you don’t need add or commit. Just:

git push upstream.

If you get conflicts, don’t worry.

1) git fetch upstream

2) git merge upstream/master

Hopefully your conflicts fix themselves automatically. Add-commit-push them again, push upstream again, and you should be good.

EDIT: I’ve just done the same thing with the BigBlueButton repo, because I want to test whether changing one single word in the main source code is enough to make modular locales a reality. I’m just posting it here so I don’t forget that I called it bbb-master instead of upstream.

EDIT: Okay, not nearly as confident about the whole “change one word” thing anymore. Ah well, experimentation is the name of the game,

Another case of “He sed, she sed.”

Justin — Thu, 03 May 2012 20:42:36 +0000

I’m writing a series of installation scripts for BigBlueButton to automate the various files where we need to insert text during the installation of the Polling Module, and I’m having a hell of a time with sed.

Here’s what’s happening: I’ve got this node in build.xml

depends=”build-bbb-main, build-bbb-main-test, build-chat, build-viewers, build-listeners, build-present, build-breakout”
description=”Compile main, chat, viewers, listeners, present, breakout modules”>

and I need to add build-polling to the depends line, and polling to the description. I’ve already got this bit of scripting to see if the polling references are already there, read the entire node, and insert the polling bits right after the mentions of main. Here’s how it looks:

dependsLine=`grep -A2 ‘ if echo “$dependsLine” | grep -q -A2 “polling”; then
echo “build-polling already included in build-main-chat-viewers-listeners-present. Continuing with installation.”
else
replaceLine=${dependsLine/”Compile main, “/”Compile main, polling, “}
replaceLine=${replaceLine/”build-bbb-main, “/”build-bbb-main, build-polling, “}
sed -i ‘s/ $dependsLine / $replaceLine /g’ /tmp/file.txt
fi

(I’m using a temporary file instead of build.xml just for testing.)

The trouble is, that last sed line is not working at all. No matter how I try it, it either throws errors or doesn’t do anything. The weird thing is, I already have another script where I replace some text with the contents of a $variable with sed. And yes, dependsLine and replaceLine look the way they should.

EDIT: Also, there are no special characters to escape in either string.

EDIT: To heck with it. From now on, any node that includes build-chat will also include build-polling, once you run these installation scripts. I was trying to do it a more elegant way, but that’s just not in the cards.

Justin — Thu, 19 Apr 2012 20:59:50 +0000

Short version: An installation of BigBlueButton + Polling Module on a remote server will intermittently, and to all appearances randomly, throw exceptions and experience Shared Object failures before the user even has a chance to do anything. Arg.

HellDAP

Justin — Wed, 28 Mar 2012 19:23:47 +0000

I am getting nowhere on this LDAP authentication thing. I’ve got some code that behaves like a phone book; you enter a username, you get some information. What I NEED is actual authentication: put in your username and password, it looks at the information that you’ve just unlocked and decides what it’s going to do with you. But every example I find looks like the exact same thing I’m doing! I either need more information about the actual Seneca LDAP server, or I need an entirely new way to go about getting this information.

EDIT: It does not -expletive deleted- help when every example I find does things slightly differently. There are any number of tiny variations and combinations of variations that could be the root of the problem, but until I get the information I need all I can do is blunder around and hope I trip over the right one. Grrrrr.

Too weird and aggravating for a clever title

Justin — Mon, 26 Mar 2012 19:53:43 +0000

So I try to get into my BBB development environment, and not only is Red5 gone but also the terminal no longer recognizes bbb-conf as a command. What the -mild profanity- is that -slightly stronger profanity-??? Now I’m having to screw around with GitHub to make sure everything is back to where it was before the weekend. Grr.

EDIT: Alright, things are back to normal on the development ranch. I still have no idea what happened in the first place, but whatever.

EDIT: I spoke too soon. My installation, for whatever odd reason, seems to be borked. Looks like a job for reinstalling the whole OS.

Just kidding, I’ll post on the setup forum (put down the heart medication, Fred)

EDIT: I’ve discovered the cure to all BigBlueButton-related problems! Simply blog about your intentions to share the problem with the development community, and make mention of wiping the OS. After that, your installation will work fine.