Automatically configure local IP address for iOS device development

[aleclarson]

Fixes#6291, #4351, #4245, #1597, and #883.

It's worth mentioning that I did not change the call to /status because it's in a "Run Script" phase. Any ideas on fixing that as well?

[facebook-github-bot]

By analyzing the blame information on this pull request, we identified @foghina, @nicklockwood and @javache to be potential reviewers.

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[javache]

I'm not a huge fan of doing this automatically, as at some point, it will be very confusing where this IP is coming from, the wrong IP will be used, or people won't find the magic shell scripts embedded in the xcode project.

[Kureev]

@aleclarson looks nice 👍

[aleclarson]

@javache

1. Is it worth moving the shell script to a .sh file for easier finding?
2. Maybe we could have a manually set IP that overrides the automatically set IP?
3. Necessary fixes to the shell script will appear in the issues over time. 😄

[Kureev]

@aleclarson

Maybe we could have a manually set IP that overrides the automatically set IP?

I think it should be something like an optional parameter: react-native run-ios --host=192.168.1.10

[aleclarson]

@Kureev Ooh I really like that.

We could also take advantage of NSUserDefaults just like the port uses [standardDefaults integerForKey:@"websocket-executor-port"].

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[aleclarson]

The recent changes include:

• The script that starts the React Packager now uses the automatic IP address.
• Combined the IP script and Packager script into local-cli/start_packager.sh.

Changes still needed:

• Add --host and --port flags to the run-ios command.
• Somehow access the custom host and port inside the start_packager.sh script.

[mroswald]

This depends strongly on your system (only macbooks on wifi, right?). And if you assume that you're on a mac, you could just use ipconfig getifaddr en0

[aleclarson]

@mroswald I just copied that line from a commit linked by this comment. This has also been discussed in depth in #4245. It can't be ipconfig getifaddr en0 because you might change networks often (eg: you go to a coffee shop, or you use ethernet at work). I doubt what I have is the perfect solution; but as long as there is a way to override the automatic IP, we can use the current solution. If anyone finds a way to make it more robust, just submit a PR.

[frantic]

I still don't fully understand what happens when you have, say, en0 and en1? Does it pick the random one?

[aleclarson]

Sorry, I haven't tested that scenario. Maybe @christopherdro knows.

[christopherdro]

@frantic has a good point that this might not work for all cases.

I believe new macs that are only wireless use en0 however, older macs that use en0 when wired in and en1 for wifi.

[frantic]

ipconfig getifaddr en0 is less magical and will be easier to fix in the future. If we want to go the more complex route, we can use ifconfig command to find and store all available interfaces, and somehow offer the user to choose from the list. I don't think it's worth it at this point.

[frantic]

Slightly different approach for addressing the same problem: #6345

I guess the main difference is that in #6345 we store the IP address in a file inside the main app's bundle, but here we modify a header file. I guess both have pro's and con's. In Facebook's case we have a copy of RN inside the repo, which means every Xcode build will change the header with the IP and might cause troubles. Using a text file is nice when we consider a potential use case of downloading pre-built binaries (instead of compiling the shell ourselves), in which case it's much easier to swap the IP.

I like that RCTServerUtils also abstracts away the port as well (if we decide to change it at some point).

@javache good point, but I think it's worth it. Now basically everyone has to mess with the IP setting every time they want to test on device. We can solve the 80%-case and have clearer error messages on how to deal with the rest 20%.

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[aleclarson]

Weird. I haven't updated the pull request in 6 days, but the bot keeps saying I'm updating it. 😯

[facebook-github-bot]

@aleclarson updated the pull request.

[facebook-github-bot]

@aleclarson updated the pull request.

[ghost]

@foghina would you mind taking a look at this pull request? It's been a while since the last commit was reviewed.

[christopherdro]

Sorry for the late response.

This is what I've been using internally for the last couple months and has been serving us well.

wildlifela@a98fc4d

Then update your AppDelegate.m.

#import"RCTIPAddress.h"
#if RCT_DEV
#if TARGET_OS_SIMULATOR
jsCodeLocation = [NSURLURLWithString:@"http://localhost:8081/index.ios.bundle?platform=ios"];
#elseNSString *URLString = [NSStringstringWithFormat:@"http://%@:8081/index.ios.bundle?platform=ios", LOCAL_IP_ADDRESS ];
jsCodeLocation = [NSURLURLWithString:URLString];
#endif
#else
jsCodeLocation = [[NSBundlemainBundle] URLForResource:@"main"withExtension:@"jsbundle"];
#endif

[rauchy]

How about https://www.npmjs.com/package/react-native-network-boot ?

[ghost]

@aleclarson do you have any updates for this pull request? It's been a while since the last update so wanted to check in and see if you've looked at the requested changes.

[ghost]

@aleclarson updated the pull request.

[aleclarson]

I noticed RCTDevMenu has a workaround of its own:

- (NSURL *)packagerURL
{
NSString *host = [_bridge.bundleURL host];
if (!host) {
returnnil;
}
NSString *scheme = [_bridge.bundleURL scheme];
NSNumber *port = [_bridge.bundleURL port];
if (!port) {
port = @8081; // Packager default port
}
return [NSURLURLWithString:[NSStringstringWithFormat:@"%@://%@:%@/message?role=shell", scheme, host, port]];
}

If RCTPackagerUtils had access to the RCTBridge, we could just go with that solution.

Otherwise, we can just replace that code with [RCTPackagerUtils URLForPath:].

[facebook-github-bot]

@aleclarson updated the pull request.

[javache]

Hi @aleclarson, @nathanajah is working on open-sourcing our internal component that provides automatic packager discovery and will incorporate your changes. Thank you for your work on this PR!

[aleclarson]

@javache Sounds good!

[isaaclem]

Is there any recommended way to handle on running ios devices?