Modules #1177

snapwich · 2017-05-04T19:29:03Z

Type of change

Feature
Refactoring (no functional changes, no api changes)
Build related changes

Description of change

The module system as described in #1086

For an example of converting an existing adapter to a module look at the adaption in file src/adapters/rubicon.js to modules/rubiconAdapter.js and tests from test/spec/adapters/rubicon_spec.js to test/spec/modules/rubiconAdapter_spec.js.

New express module was added as well. With the new express module and the conversion of the rubcion adapter to a module the build currently spits out 3 files: prebid.js, express.js, and rubiconAdapter.js. These can then being bundled for usage after gulp is run using gulp bundle --modules=express,rubiconAdapter. The --modules flag is optional if you just want to bundle all modules found.

snapwich · 2017-05-08T19:26:15Z

Will update this pull-request with external module support using require.resolve for npm support.

protonate

As discussed let's incorporate rollup and plugins to achieve this functionality while preserving & standardizing on es6 module syntax, also to minimize the required changes to adapters, tests, etc.

snapwich · 2017-05-17T19:25:10Z

I've added a commit (5d7ade4) showing an example of splitting modules only using webpack code-splitting. I didn't update the express module as we'll still need to figure out the best way to get config to modules, but the rubiconAdapter module has been updated.

snapwich · 2017-06-07T00:48:13Z

I've updated this pull-request with the rubicon, appnexus, and appnexusAst adapters converted to modules.

To test you can clone the repo and checkout the modules branch, yarn install, and then gulp which generates the built components. Follow that with gulp bundle which bundles all of the built modules (by default) into an individual file build/dist/bundle.js.

You can also use gulp bundle --modules=appnexusBidAdapter,appnexusAstBidAdapter to specify only certain modules to be included in the bundle.js file.

gulp build will do both the build step and the bundle step as a convenience and accepts the same --modules argument.

Besides searching for modules in the ./modules directory, modules are also sought in node_modules directory from within the repo, or node_modules directories in parent folders. You can specify external modules by name using the same --modules argument. e.g. gulp build --modules=someNpmPkg. These external modules are included in the build process. i.e. external modules are ran through the webpack build so they can import, use babel features, and be written as a regular module would be written from within the project.

Starting tomorrow I will begin the process of converting the rest of the adapters into modules.

Note: One final thing probably needed before being production ready is to create the webpack plugin to exclude the manifest and jsonp loading code from the webpack runtime. These bundles work fine as is implemented here, but excluding that code will reduce the bundle size.

snapwich · 2017-06-07T18:44:38Z

Another thing to note: in order to support supportedMediaTypes that was handled by the adapter-loader, video-loader, and native-loader; I have added an additional parameter to adaptermanger.registerBidAdapter that supports a configuration object that includes supportedMediaTypes

protonate · 2017-06-07T21:08:22Z

@matthewlane @jaiminpanchal27 @dbemiller added all of you as reviewers, please have a look as this is working towards final and will need to be understood by all of us.

ready for re-review

dbemiller · 2017-06-08T20:40:24Z

I looked this over... but I must admit, I think I'm too late to the party to give much serious feedback.

A build task which creates a bundle with only the adapters you need looks like a great contribution. The spec was also super clear and well-written.

I'm getting hung up on the claim: "we want dynamic runtime inclusion without async loading." The main benefit of for dynamic runtime inclusion which I can think of is the fact that you can download only the parts of the javascript which you need. Since it sounds like you still expect publishers to have downloaded it all, then what benefit is there to loading things at runtime?

On the other hand... I don't actually see anything in the diff which adds APIs to allow that. So am I missing something, or did that part of the spec get dropped?

The rest is mostly details. Will leave some comments on individual files about them.

dbemiller · 2017-06-08T20:45:28Z

modules/express.js

@@ -0,0 +1,200 @@
+


Could you leave comments at the top of new files describing (at a high level) what their purpose is?

It's helpful for a few reasons... but those are even more pronounced now that we're making Modules.

It helps newcomers get their bearings on the codebase

It helps users decide whether or not they should include the module (at a glance) in their bundle

It (theoretically) improves code quality over time, because some future engineer tasked with implementing "feature X" can tell at a glance which (if any) code makes sense in this module.

dbemiller · 2017-06-08T20:46:48Z

package.json

    "gulp-header": "^1.7.1",
+    "gulp-if": "^2.0.2",
+    "gulp-jscs": "^3.0.2",


This looks accidental... are you using jscs anywhere? I thought we got rid of that with the move to eslint.

I added the gulp-if but am not sure what gulp-jscss deal is. If it was removed I might have accidentally re-added it after a merge conflict, I can remove it if its no longer needed.

yeah, it was removed here in #1111

dbemiller · 2017-06-08T21:02:52Z

src/adapter.js

@@ -0,0 +1,24 @@
+function Adapter(code) {


Does this replace src/adapters/adapter.js? I don't see that file being deleted... but they're duplicates.

yes, this replaces src/adpaters/adapter.js. I haven't removed the old file yet because I haven't finished converting all the adapters to modules yet, so the non-converted adapters still look for this file. As soon as I'm done converting the adapters to modules I'll make sure this is moved properly in git.

dbemiller · 2017-06-08T21:07:46Z

src/config.js

+    configHandlers[property] = [];
+  }
+  configHandlers[property].push(handler);
+}


What happens if getConfig is called after the last call to setConfig? If I were calling these APIs, I would still expect my callback to be executed.

Actually... a related problem: I would expect my callback to be executed exactly once. It doesn't affect your use-cases here... but looking at this file in isolation, I'd still consider it a bug.

Related: are unit tests coming soon?

dbemiller · 2017-06-08T21:19:21Z

modules/rubiconBidAdapter.js

@@ -22,6 +24,10 @@ const VIDEO_ENDPOINT = '//fastlane-adv.rubiconproject.com/v1/auction/video';

 const TIMEOUT_BUFFER = 500;

+var ACCOUNT_ID;
+getConfig('rubiconAdapter', config => ACCOUNT_ID = config.accountId);


this looks..... well, very dangerous. By setting a variable in an async callback, you've inherently created a race condition. This works (I would guess?) because a user will (should?) call setConfig sometime between the time that this code runs, and when they kick off an auction... but if someone tried to use ACCOUNT_ID in this file outside of those auction-related functions, you could have some really tricky bugs to track down.

Part of me wants to say "your adapter, your business :p" but.... given that you're prototyping an architecture here, I'm a bit concerned because I can't really see a way to do it safely.

It seems to me like it would have to look something like:

getConfig('myNamespace', config => { var property = config.myParam; module.exports = stuffWhichUses(property); });

...which I'm pretty sure modules don't actually let you do.

Can you write an example of a very small module which reads a value from the config, and then exports something useful once the param has been read? Without assuming that callers follow that implicit import files, setConfig, callUsefulMethod order.

I don't know if I'd call it a race condition, but there is definitely a dependency problem. However, I didn't create it. The problem is inherent in the way we execute pbjs.que blocks and when module code is ran by webpack. I detailed this in a diagram I pasted to slack awhile back to describe our problem:

We effectively have a circular dependency based on the current architecture of Prebid.js. This isn't a new problem, it's just now more obvious because this is the first time someone has attempted to apply global config (config across adUnits and auctions) to an adapter (or in this case now, a module). Currently configuration can only come in through bid.params which makes this condition less obvious, but is also not an ideal solution for handling global config.

Using a callback was the solution I found to solve this. By registering a callback that sets the config it allows us to defer configuration until after setConfig is called. I think the dependency is intuitive (that you must setConfig for a module before you can use that config), so I'm not too worried about the implications.

I originally tried to solve the problem with promises, but then I ran into an actual race condition when the scheduled config callback wasn't called until after the auction had ran.

It's possible to not register the adapter to Prebid.js until the config has been received (the registration being the real export that Prebid cares about, module.exports is only used for testing), but then you're just swapping a set of config error messages, for a more cryptic and general error message. i.e. "Rubcion Adapter missing required config Account ID" vs "Rubicon Adapter was not found".

Yeah, that makes sense. I also agree that there's an architectural problem with prebid right now, and your diagram distills it down well.

One obvious fix that jumps out at me is if we moved towards an API like requestBids(config) instead, and then implement write requestBids to forward the config through to any adapters. That would also move us one step closer to multiple auctions, since it's clear what the config for each particular auction is.

I know fixing all that is outside the scope of this PR... but I wonder if we could get things a little closer to that ideal, without too much effort?

For example, right now your modules look like:

getConfig(); adapter = makeAdapter(config); adaptermanager.registerBidAdapter(adapter, ...);

Instead, we could do something like:

function adapterLoader(config) { return makeAdapter(config); } adaptermanager.registerBidAdapterLoader(adapterLoader, ...);

Then load the config inside registerBidAdapterLoader, and forward the call through to registerBidAdapter in the config callback.

Main benefits being:

The circular dependence stays within prebid-core, and out of the modules

Adapters would be easier to unit test (can send through mock configs)

Another benefit (disjoint from the first... so take it as an independent suggestion) is that you could namespace the global config. For example, if the bidder code is rubicon, then registerAdapterLoader function could decide (universally) that rubiconAdapter is where your config data goes. Then when it actually executes your load function, it can send that data right back to you.

Benefits there being:

Lower publisher learning curve (config naming conventions consistent across all adapters)

Adapters can't trample each others' config values accidentally

snapwich · 2017-06-08T22:24:48Z

So am I missing something, or did that part of the spec get dropped?

@dbemiller I wrote a longer response, but for whatever reason github posted it twice and when I deleted the extra one github decided to delete both of them.... It might still be in your e-mail, but if not, here's the gist of it:

The requirement of dynamic runtime inclusion was modified in a follow-up meeting concerning modules that happened after the spec was written. It was stressed that in an effort to keep es6 module syntax we should get rid of the runtime requirement. We still want quick dynamic inclusion though. And by that I mean, concatenating built files together to include required functionality (a process which is fast and could be part of a server-side request), or including them individually like:

<script src="./prebid.js"></script>
<script src="./rubiconBidAdapter.js"></script>
<script src="./appnexusBidAdapter.js"></script>
<script>
pbjs.processQueue();
pbjs.que.push(function() {
  ... stuff here
});
</script>

GLStephen · 2017-06-08T22:36:13Z

@snapwich The latter (individual scripts) approach would likely see benefits in maintenance and performance in an HTTP2 environment if managed properly. In terms of caching and config and in terms of delivering from a central CDN while also allowing modularity with a known, off the shelf for everyone, mechanism: HTML.

snapwich · 2017-06-08T22:43:24Z

@GLStephen I think it's a feasible approach. However, I don't think it'll be used often due to the rampant use of tag managers within the ad space and the hard requirement that the prebid.js code (that contains the webpack runtime) be loaded first (so other modules have something to register with).

For that reason I think the server-side concatenation of the bundle will be more common. I think that is still a very useful addition to Prebid.js though. It'll allow us to do things like:

<script src="./prebid.js?modules=rubiconBidAdapter,appnexusBidAdapter"></script>

and get exact bundles of only the code we want on request.

dbemiller · 2017-06-09T14:06:05Z

I agree that server-side or build-time concatenation generally makes much more sense than multiple script tags on the page.

If you do want to leave that pattern open as a possibility, then the relevant parts of each module in this code (e.g. the registerBidAdapter code) should be wrapped inside the cmd. Otherwise you'll be forcing people to load prebid-core synchronously.

But that's pretty convoluted and icky if loading individual modules is a non-goal.

protonate

LGTM!

navneetpandey · 2017-06-27T18:21:31Z

Please note that this document http://prebid.org/dev-docs/bidder-adaptor.html now contains many broken link.

bretg · 2017-06-27T19:55:32Z

Thanks @navneetpandey - covered in prebid/prebid.github.io#275. May have missed some -- review welcome.

navneetpandey · 2017-06-30T11:21:57Z

Seem consistent now @bretg . Thanks.

* initial implementation of module system * express module * conversion of rubicon adapter to module system * use var instead of let in node scripts * example of using modules for concatenation through webpack code-splitting * fixes for webpack code-splitting * changes including config through setConfig and getConfig callback * external modules working, but node_modules in project still needs babel-laoder fix * gulp watch bundle building * fix for babel-loader node_modules exclusion and bundle only * updated express module to work in new webpack module format * gulp default, don't bundle * updated rubicon adapter to latest * added appnexus and appnexusAst to modules. added support for native and video registering * added logging for files getting bundled * fixed bug in bundling external modules * aardvark bid adapter converted to module * upadte appnexus bid adpater names and get ride of mocha loader tests * adblade bid adapter converted to module * adblund bid adapter converted to module * adbutler bid adapter converted to module * adequant bid adapter converted to module * adform bid adapter converted to module * adkernel bid adapter converted to module * admedia bid adapter converted to module * admixer bid adapter converted to module * fix pbjs global for tests so load order doesn't matter * adsuppply bid adpater converted to module * use $$PREBID_GLOBAL$$ rather than pbjs in test files * adyoulike bid adapter converted to module * fixed adapter location in some adapters * aol bid adapter converted to module * atomx bid adapter converted to module * audienceNetwork bid adapter converted to module * beachfront bid adapter converted to module * bidfluence bid adapter converted to module * brightcom bid adapter converted to module * centro bid adapter converted to module * rubicon spec file name changed * conversant bid adapter converted to module * criteo bid adapter converted to module * districtmDMX bid adapter converted to module * fidelity bid adapter converted to module * getintent bid adapter converted to module * gumgum bid adapter converted to module * hiromedia bid adapter converted to module * huddledmasses bid adapter converted to module * indexExchange bid adapter converted to module * inneractive bid adapter converted to module * innity bid adapter converted to module * jcm bid adapter converted to module * komoona bid adapter converted to module * kruxlink bid adapter converted to module * lifestreet bid adapter converted to module * mantis bid adapter converted to module * memeglobal bid adapter converted to module * nginad bid adapter converted to module * openx bid adapter converted to module * piximedia bid adapter converted to module * prebidServer bid adapter converted to module * pubgears bid adapter converted to module * pubmatic bid adapter converted to module * pulsepoint and pulsepointLite bid adapter converted to module * quantcast bid adapter converted to module * rhythmone bid adapter converted to module * added adkernel alias to headbidding to adkernalBidAdapter module * roxot bid adapter converted to module * sekindoUM bid adapter converted to module * remove baseAdapter * serverbid bid adapter converted to module * sharethrough bid adapter converted to module * smartyads bid adapter converted to module * update conversantBidAdapter to support video * get rubiconBidAdapter latest * move rubicon adapter to module properly * sonobi bid adapter converted to module * sovrn bid adapter converted to module * springserve bid adpater converted to module * stickyadstv bid adapter converted to module (and removed console logs) * convert tapsense bid adapter to module * thoughtleadr bid adapter converted to module * trion bid adapter converted to module * triplelift bid adapter converted to module * twenga bid adapter converted to module * underdogmedia bid adapter converted to module * vertamedia bid adapter converted to module * vertoz bid adapter converted to module * wideorbit bid adapter converted to module * widespace bid adapter converted to module * xhb bid adapter converted to module * yieldbot converted from bid adapter to module * smartadserver bid adapter converted to module * remove old adapter class in adapter folder * converted analytics adapters to modules * removed loaders from build * added plugin to remove file manifest and jsonp code from webpack runtime * cox bid adapter converted to module * carambola bid adapter converted to module * eplanning bid adapter converted to module * --modules arg allows .json file as parameter * removing config module for now * don't bundle AnalyticsAdapter.js in common chunk * unruly bid adapter converted to module * fixed case-sensitivity in including webpack plugin * can't use class keyword, updated to constructor function for plugin * added jsdoc to express and new webpack plugin * fixed sourcemaps for concat step in bundle * use external sourcemap for bundle instead of inline * rename prebid.js to prebid-core.js and main bundle outputs prebid.js now * updated README.md to be module specific * added filter to ignore hidden files in modules directory

jeremiegirault · 2017-11-02T17:21:26Z

Hi, as of this day, the RequireEnsureWithoutJsonp.js has no effect. The PR #1785 references the chunk function name for some reason, but I don't think it works as it should. Worth fixing ?

snapwich · 2017-11-08T00:17:56Z

@jeremiegirault what do you mean, it has no effect? I just pulled the latest and built prebid and it still seems to be successfully stubbing the jsonp loader code.

jeremiegirault · 2017-11-08T09:50:05Z

Hi, In my tests I had troubles and the plugin was not active indeed. But after using the good tag/master looking at the generated code it indeed seem to replace the requireEnsure function :

/******/ 	__webpack_require__.e = function requireEnsure(chunkId) {
/******/ 		if(installedChunks[chunkId] === 0)
/******/ 		  return callback.call(null, __webpack_require__);
/******/ 		else
/******/ 		  console.error('webpack chunk not found and jsonp disabled');
/******/ 	};

However callback does not exists. Indeed the prototype of requireEnsure used to be function requireEnsure(chunkId, callback). Now it returns a Promise (probably due to some internal webpack changes?).

Looking at the source without the plugin :

/******/ 	__webpack_require__.e = function requireEnsure(chunkId) {
/******/ 		var installedChunkData = installedChunks[chunkId];
/******/ 		if(installedChunkData === 0) {
/******/ 			return new Promise(function(resolve) { resolve(); });
/******/ 		}
/******/

...

So I question the usefulness of this webpack plugin which seem outdated in the context ?

snapwich · 2017-11-08T23:37:54Z

@jeremiegirault yeah that code is probably useless. It was an attempt to fail nicely if JSONP loading was attempted, but I don't think it's ever running. I added a pull-request to just remove that code block entirely.

* initial implementation of module system * express module * conversion of rubicon adapter to module system * use var instead of let in node scripts * example of using modules for concatenation through webpack code-splitting * fixes for webpack code-splitting * changes including config through setConfig and getConfig callback * external modules working, but node_modules in project still needs babel-laoder fix * gulp watch bundle building * fix for babel-loader node_modules exclusion and bundle only * updated express module to work in new webpack module format * gulp default, don't bundle * updated rubicon adapter to latest * added appnexus and appnexusAst to modules. added support for native and video registering * added logging for files getting bundled * fixed bug in bundling external modules * aardvark bid adapter converted to module * upadte appnexus bid adpater names and get ride of mocha loader tests * adblade bid adapter converted to module * adblund bid adapter converted to module * adbutler bid adapter converted to module * adequant bid adapter converted to module * adform bid adapter converted to module * adkernel bid adapter converted to module * admedia bid adapter converted to module * admixer bid adapter converted to module * fix pbjs global for tests so load order doesn't matter * adsuppply bid adpater converted to module * use $$PREBID_GLOBAL$$ rather than pbjs in test files * adyoulike bid adapter converted to module * fixed adapter location in some adapters * aol bid adapter converted to module * atomx bid adapter converted to module * audienceNetwork bid adapter converted to module * beachfront bid adapter converted to module * bidfluence bid adapter converted to module * brightcom bid adapter converted to module * centro bid adapter converted to module * rubicon spec file name changed * conversant bid adapter converted to module * criteo bid adapter converted to module * districtmDMX bid adapter converted to module * fidelity bid adapter converted to module * getintent bid adapter converted to module * gumgum bid adapter converted to module * hiromedia bid adapter converted to module * huddledmasses bid adapter converted to module * indexExchange bid adapter converted to module * inneractive bid adapter converted to module * innity bid adapter converted to module * jcm bid adapter converted to module * komoona bid adapter converted to module * kruxlink bid adapter converted to module * lifestreet bid adapter converted to module * mantis bid adapter converted to module * memeglobal bid adapter converted to module * nginad bid adapter converted to module * openx bid adapter converted to module * piximedia bid adapter converted to module * prebidServer bid adapter converted to module * pubgears bid adapter converted to module * pubmatic bid adapter converted to module * pulsepoint and pulsepointLite bid adapter converted to module * quantcast bid adapter converted to module * rhythmone bid adapter converted to module * added adkernel alias to headbidding to adkernalBidAdapter module * roxot bid adapter converted to module * sekindoUM bid adapter converted to module * remove baseAdapter * serverbid bid adapter converted to module * sharethrough bid adapter converted to module * smartyads bid adapter converted to module * update conversantBidAdapter to support video * get rubiconBidAdapter latest * move rubicon adapter to module properly * sonobi bid adapter converted to module * sovrn bid adapter converted to module * springserve bid adpater converted to module * stickyadstv bid adapter converted to module (and removed console logs) * convert tapsense bid adapter to module * thoughtleadr bid adapter converted to module * trion bid adapter converted to module * triplelift bid adapter converted to module * twenga bid adapter converted to module * underdogmedia bid adapter converted to module * vertamedia bid adapter converted to module * vertoz bid adapter converted to module * wideorbit bid adapter converted to module * widespace bid adapter converted to module * xhb bid adapter converted to module * yieldbot converted from bid adapter to module * smartadserver bid adapter converted to module * remove old adapter class in adapter folder * converted analytics adapters to modules * removed loaders from build * added plugin to remove file manifest and jsonp code from webpack runtime * cox bid adapter converted to module * carambola bid adapter converted to module * eplanning bid adapter converted to module * --modules arg allows .json file as parameter * removing config module for now * don't bundle AnalyticsAdapter.js in common chunk * unruly bid adapter converted to module * fixed case-sensitivity in including webpack plugin * can't use class keyword, updated to constructor function for plugin * added jsdoc to express and new webpack plugin * fixed sourcemaps for concat step in bundle * use external sourcemap for bundle instead of inline * rename prebid.js to prebid-core.js and main bundle outputs prebid.js now * updated README.md to be module specific * added filter to ignore hidden files in modules directory

mkendall07 added the in progress label May 4, 2017

snapwich requested a review from mkendall07 May 4, 2017 19:34

protonate self-requested a review May 8, 2017 18:23

protonate assigned protonate and snapwich and unassigned protonate May 8, 2017

protonate previously requested changes May 17, 2017

View reviewed changes

protonate added the needs update label May 17, 2017

mkendall07 mentioned this pull request May 23, 2017

Structure proposal #1222

Closed

protonate added this to the Prebid 0.25.0 milestone Jun 2, 2017

snapwich force-pushed the modules branch 2 times, most recently from 28ce904 to 16a9f4d Compare June 6, 2017 22:08

ndhimehta mentioned this pull request Jun 7, 2017

Prebid 1.0 - module interface #1086

Closed

protonate added needs review and removed needs update labels Jun 7, 2017

protonate requested review from matthewlane, jaiminpanchal27 and dbemiller June 7, 2017 21:07

dbemiller suggested changes Jun 8, 2017

View reviewed changes

snapwich force-pushed the modules branch from c6faf92 to a2c5c1f Compare June 23, 2017 21:19

snapwich added 2 commits June 23, 2017 14:25

use external sourcemap for bundle instead of inline

1dbd41d

rename prebid.js to prebid-core.js and main bundle outputs prebid.js now

50e9bff

protonate approved these changes Jun 26, 2017

View reviewed changes

snapwich added 2 commits June 26, 2017 14:42

updated README.md to be module specific

1e2d6e5

added filter to ignore hidden files in modules directory

15e8565

mkendall07 merged commit f6449d6 into prebid:master Jun 26, 2017

mkendall07 removed the in progress label Jun 26, 2017

protonate removed the needs review label Jun 27, 2017

rmloveland mentioned this pull request Jun 27, 2017

Update all mentions of adapters.json to reference new adapter "module" system prebid/prebid.github.io#277

Closed

jaiminpanchal27 mentioned this pull request Jun 27, 2017

Add ucfunnel adapter #1263

Merged

9 tasks

dbemiller mentioned this pull request Jun 27, 2017

Audience Network: Add support for video format #1292

Merged

8 tasks

elljoh mentioned this pull request Jun 28, 2017

Yieldbot refresh - nextPageview with specific refresh slots #1281

Merged

1 task

This was referenced Jun 29, 2017

Gulp serve shouldn't rebuild on integrationExamples #1333

Closed

Gulp serve doesn't rebuild bundle #1334

Closed

dbemiller mentioned this pull request Jul 12, 2017

Sharethrough Adapter - set bidderCode on invalid bid #1295

Merged

4 tasks

mkendall07 mentioned this pull request Jul 13, 2017

Correct way to build with Analytics? #1379

Closed

dbemiller mentioned this pull request Aug 10, 2017

Question: When will prebid 0.26.1 be available on http://prebid.org/download.html ? #1430

Closed

snapwich mentioned this pull request Nov 8, 2017

just remove require.ensure entirely #1816

Merged

1 task

robertrmartinez deleted the modules branch July 5, 2023 19:45

Modules #1177

Modules #1177

Uh oh!

Conversation

snapwich commented May 4, 2017

Type of change

Description of change

Uh oh!

snapwich commented May 8, 2017

Uh oh!

protonate left a comment

Choose a reason for hiding this comment

Uh oh!

snapwich commented May 17, 2017

Uh oh!

snapwich commented Jun 7, 2017

Uh oh!

snapwich commented Jun 7, 2017

Uh oh!

protonate commented Jun 7, 2017

Uh oh!

dbemiller commented Jun 8, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

snapwich Jun 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

dbemiller Jun 9, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

snapwich commented Jun 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

GLStephen commented Jun 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

snapwich commented Jun 8, 2017

Uh oh!

dbemiller commented Jun 9, 2017

Uh oh!

protonate left a comment

Choose a reason for hiding this comment

Uh oh!

navneetpandey commented Jun 27, 2017

Uh oh!

bretg commented Jun 27, 2017

Uh oh!

navneetpandey commented Jun 30, 2017

Uh oh!

jeremiegirault commented Nov 2, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

snapwich commented Nov 8, 2017

Uh oh!

jeremiegirault commented Nov 8, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

snapwich commented Nov 8, 2017

Uh oh!

Uh oh!

snapwich Jun 8, 2017 •

edited

Loading

dbemiller Jun 9, 2017 •

edited

Loading

snapwich commented Jun 8, 2017 •

edited

Loading

GLStephen commented Jun 8, 2017 •

edited

Loading

jeremiegirault commented Nov 2, 2017 •

edited

Loading

jeremiegirault commented Nov 8, 2017 •

edited

Loading