One to many repository maps

[erichoracek]

Carthage allows one repository to have many framework targets. When a carthage build is performed, it builds them all. An example of this is CocoaLumberjack, which produces two iOS frameworks from a single repository:

carthage build CocoaLumberjack --platform ios
*** Building scheme "CocoaLumberjack-iOS" in Lumberjack.xcworkspace
*** Building scheme "CocoaLumberjackSwift-iOS" in Lumberjack.xcworkspace

this produces:

Carthage/Build/iOS/CocoaLumberjack.framework
Carthage/Build/iOS/CocoaLumberjack.framework.dSYM
Carthage/Build/iOS/CocoaLumberjackSwift.framework
Carthage/Build/iOS/CocoaLumberjackSwift.framework.dSYM

It seems as if the [RepositoryMap] should allow for a single repository to map to many framework references. You could imagine it would look something like this:

[RepositoryMap]
  CocoaLumberjack CocoaLumberjack CocoaLumberjackSwift

[tmspzz]

How do you see this working when:

• rome (upload | download) CocoaLumberjack ? Should rome try to upload/download all resolved names?
• rome list ([--missing] | [--present]) ?
• Should rome report cache hits only when ALL resolved names are present?
• Should rome report misses when ANY of the resolved names is a miss?

[erichoracek]

I think this one is a bit different from dSYMS. If you have specified that a framework has a one-to-many repository/framework relationship, it would seem like rome would try to upload/download all names if they're specified, and consider any operation a failure if one of the frameworks are unavailable. As far as I know, there is no way to tell carthage to only build a specific subset of frameworks from a repo.

In terms of the above questions, this would result in:

Should rome try to upload/download all resolved names?

Yes, and consider the operation a failure if any are missing.

Should rome report cache hits only when ALL resolved names are present?

Yes

Should rome report misses when ANY of the resolved names is a miss?

Yes

Unfortunately, since Rome isn't able to communicate with the internals of Carthage, it's not as simple as just parsing the output of xcodebuild -list like carthage:

https://github.com/Carthage/Carthage/blob/master/Source/CarthageKit/Xcode.swift#L167-L192
https://github.com/Carthage/Carthage/blob/master/Source/CarthageKit/Xcode.swift#L745-L769

[tmspzz]

@erichoracek Please try https://github.com/blender/Rome/releases/tag/v0.6.0.8

I think that rome list could be improved to report what alias of the framework is missing

[tmspzz]

Fixes a crash in v0.6.0.8 https://github.com/blender/Rome/releases/tag/v0.6.0.9

[erichoracek]

From trying out version 0.6.0.9, this doesn't seem to be working quite as expected. We have a repo named libPusher-Carthage, which builds the Pusher and SocketRocket frameworks.

When specifying the following [RepositoryMap]:

[RepositoryMap]
    libPusher-carthage = Pusher, SocketRocket

and this Cartfile.resolved:

github "Automatic/libPusher-carthage" "1.6.1"

I get the following behavior on rome upload:

$ rome upload  
Successfully uploaded Pusher to: Pusher/Pusher.framework-1.6.1.zip
Successfully uploaded Pusher.dSYM to: Pusher/Pusher.framework.dSYM-1.6.1.zip

And the following behavior on rome download:

$ rome download
Donwloaded: Pusher.framework-1.6.1.zip
Unzipped: Pusher.framework-1.6.1.zip
Donwloaded: Pusher.framework.dSYM-1.6.1.zip
Unzipped: Pusher.framework.dSYM-1.6.1.zip
Error downloading  SocketRocket.framework-1.6.1.zip : The specified key does not exist.
Error downloading  SocketRocket.framework.dSYM-1.6.1.zip : The specified key does not exist.

However, if I reverse the ordering of the repository map:

[RepositoryMap]
    libPusher-carthage = SocketRocket, Pusher

I get the reverse behavior:

$ rome upload  
Successfully uploaded SocketRocket to: SocketRocket/SocketRocket.framework-1.6.1.zip
Successfully uploaded SocketRocket.dSYM to: SocketRocket/SocketRocket.framework.dSYM-1.6.1.zip

It seems like this is almost there—thanks for taking a stab at this!

[tmspzz]

@erichoracek I believe your problem are additional white spaces around frameworks names. This is addressed in pre-release v6.0.10

[erichoracek]

@blender Thanks! I'll try this out.

Re: the v0.6.0.10 release, it would appear that the binary is missing from the Github attachments.

[tmspzz]

@erichoracek Sorry about that. Fixed.

[erichoracek]

Appears to be working perfectly now, thanks!