Use readlink instead of realpath in packager.sh

[dminkovsky]

Summary

realpath is not available on macOS 12.2 and 12.4. Because of this, the following error is shown when launching an RN app with XCode (which calls packager.sh via launchPackage.command):

I am running Bash 5 but realpath is also not available with zsh.

This issue was introduced in bb8ddd6#diff-6ca7c99209bdf630550bb9e2946ce8611948c5a23b32ffb25028792ef5d48b8d, which interestingly did not change launchPackage.command. There's a recent comment on that commit that confirms this issue:

bb8ddd6#commitcomment-77818917

Changelog

[iOS] [Fixed] - Use readlink instead of realpath in packager.sh

Test Plan

[analysis-bot]

Platform	Engine	Arch	Size (bytes)	Diff
android	hermes	arm64-v8a	7,823,071	+0
android	hermes	armeabi-v7a	7,216,282	+0
android	hermes	x86	8,136,194	+0
android	hermes	x86_64	8,114,049	+0
android	jsc	arm64-v8a	9,700,797	+0
android	jsc	armeabi-v7a	8,456,276	+0
android	jsc	x86	9,652,143	+0
android	jsc	x86_64	10,249,870	+0

Base commit: b66db7a
Branch: main

[analysis-bot]

Platform	Engine	Arch	Size (bytes)	Diff
ios	-	universal	n/a	--

Base commit: b66db7a
Branch: main

[facebook-github-bot]

@philIip has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[cortinico]

Thanks for sending this over @dminkovsky

realpath is not available on macOS 12.2 and 12.4

Do you have a reference? I'm on 12.3 and I effectively have it. I'm wondering why this got removed in 12.3.

Also there are other usages of realpath in other scripts as well in the codebase, we'll have to replace it as well if this is the case.

[dminkovsky]

Hey thanks for taking a look at this @cortinico.

My only reference for 12.4 is that I'm running it, and the other reporter on the breaking commit said they were running 12.2. So, I have no idea about 12.3, but realpath is, generally, to my knowledge, a common but not ubiquitous utility, not like readlink anyway. I don't think it's ever been available by default on macOS or BSD. For example: whatwg/html-build#90 (2016).

In general, though, I find this issue a bit confusing because the breaking commit (bb8ddd6) is from February 22nd, quite some time ago. Why did this only surface now? What motivated bb8ddd6? Maybe this https://stackoverflow.com/questions/3572030/bash-script-absolute-path-with-os-x?

[dminkovsky]

Maybe just make brew install coreutils a required dependency for RN on macOS?

[cortinico]

What motivated bb8ddd6?

Happy to provide a bit more context here. (Also cc @IanChilds as the original author).

It seems like our internal Buck setup needed to use realpath to resolve path of some of our paths. What happend was that a search-and-replace was executed on all the .sh files, replacing readlink with realpath.

I think we should be fine by either:

1. Updating the RN website mentioning coreutils as one of the util to install.
2. Merge this PR as it is. Specifically the file you're touching is used to kick-off Metro for RN-Tester manually and is never invoked by Buck. We might have other occurrencies of realpath in user facing scripts, so this issue migth surface again in the future.

So, I have no idea about 12.3

FYI, I'm on 12.3.1 and I do have realpath without having to install coreutils. However which realpath ends up being inside the brew folder, so I suspect I have it pulled in because of gnupg.

[danilobuerger]

I can confirm that realpath is not available per default on macOS 12.4

[cortinico]

@dminkovsky Could you rebase?

[dminkovsky]

Sure @cortinico, will be at a computer in a couple hours will let you know.

[dminkovsky]

@cortinico I rebased

[dminkovsky]

(Tried renaming my branch, which didn't go as intended)

[facebook-github-bot]

@cortinico has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[react-native-bot]

This pull request was successfully merged by @dminkovsky in 698b147.

^{When will my fix make it into a release? | Upcoming Releases}

[cortinico]

Following up here: we decided to merge this as this usage of realpath was user facing and not used in the internal infra so we were good to go. If other occurrencies of realpath are offending a user flow, we will update them as well