build: Also output commitmeta.json

[jlebon]

This is generally useful but especially so for the pkglist alone for
higher level tools displaying releases/build outputs.

Closes: #197

[jlebon]

Still testing this!

[cgwalters]

LGTM, will let someone else review though too.

[dustymabe]

Still testing this!

let us know when ready.

[jlebon]

OK, now tested!

[miabbott]

Not super familiar with the Python bindings, but I pulled the PR and it worked for me.

[miabbott]

@jlebon Do you want to make a separate PR for the rhcos branch?

[jlebon]

I can do so once the PR is merged.

[dustymabe]

LGTM. sorry about being late

I did have one question. Is a negative value for timestamp allowed/expected?

[coreos-assembler]$ /usr/lib/coreos-assembler/commitmeta_to_json ./repo/ 82748f30b8475fa16f8e879b3f71daf00b9517188a2abca1eb13340d6dc15b7e
{                                   
  "rpmostree.inputhash" : "6577bfcb0b1d9c547b5d087eb2f36290c51750ffb66721f92303df31ea8713c4",
  "coreos-assembler.config-dirty" : false,
  "rpmostree.rpmmd-repos" : [                          
    {                                     
      "id" : "fedora",                                        
      "timestamp" : 2301568498846924800                    
    },                                 
    {                                                  
      "id" : "fedora-updates",                         
      "timestamp" : -7864432448686260224                                                    
    },                                  
    {                                   
      "id" : "fedora-updates-testing",  
      "timestamp" : 358858995915554816  
    },                                         
    {                                                      
      "id" : "dustymabe-ignition",      
      "timestamp" : 4343974019603628032                          
    }                                                            
  ],               

[lucab]

Positive timestamps are off as well (not in this century). Opposite integer endianness?

[jlebon]

Yeah, that's some GVariant implementation detail. Will tweak!

[cgwalters]

Opposite integer endianness?

Yep. libostree uses BE by default to force this issue (similar rationale as ext4). We need to swap here, which...hm. Maybe it'd be easier to have rpm-ostree compose tree --output-commitmeta-json.

(Of course there's the "JS numbers are really doubles" problem but that's only going to be an issue a rather long time from now)

[jlebon]

#231

[jlebon]

@jlebon Do you want to make a separate PR for the rhcos branch?

#233