Fix :head option for Pods from HTTP source.

[neonichu]

HTTP doesn't implement checkout_options.

While specifying :head for Pods from HTTP sources doesn't really make sense, it feels like it should still work so that the author of a Podfiledoesn't have to know about the download source of each Pod he uses.

[alloy]

I don’t think it should work, as they might then (obviously incorrect anyways) wonder why it doesn’t update. I think that instead the error message should be informative. E.g. The pod 'spatialite' is downloaded from a HTTP source, which does not support the:head' option. It only applies to SCM systems.`

(Make the message a bit extra nice, I just quickly typed something ;))

[neonichu]

True, such an error message would be better. Also a bigger undertaking, because first cocoapods-downloader would need to provide that information.

[alloy]

We don’t need to modify the exception in cocoa pods-downloader (although that one could be prettier too), but instead you can do something like:

if downloader.options_specific?
  @specific_source = downloader.checkout_options
else
  raise Informative, "Better error message."
end

[neonichu]

OK, will take a stab at that.

[neonichu]

It's not that easy, as options_specific will be false for Git as well, if no explicit commit or tag is specified. We could check if checkout_options raises but that seems very unclean to me.

[alloy]

Ah good point.

We could check if checkout_options raises but that seems very unclean to me.

While it often is, I’m not sure right now what the best way would be to implement it. So how about rescuing and raising for now and filing a ticket on cocoapods-downloader to allow for nicer error handling?

begin
  # do work
rescue RuntimeError => e
  if e.message == 'Abstract method'
    raise Informative, "Does not support #{options}"
  else
    raise
  end
end

[neonichu]

OK, makes sense. In addition to being unclean, it has the disadvantage that it will first download, then fail :)

[alloy]

Very true, the more reason we should fix this properly. But I do believe that the time it takes for a user to download an archive is nothing compared to the rage unleashed by having to debug ;)

[alloy]

Sorry for not making this note earlier, but the change also needs an accompanying test case :) Let me know if you need assistance with that.

[neonichu]

Of course I do :). Have added one, probably bad, though, because it will actually hit HTTP?

[alloy]

Thanks! When I get back, later today, I will update the test to stub it so it doesn't actually try to download.

And finally, can you also add a CHANGELOG entry to credit yourself?

[neonichu]

Updated CHANGELOG in #1972

[neonichu]

OK, figured out how to do that myself :)

[alloy]

Perfect, thanks! (I amended the patch slightly because it was clashing with another change that cleans a dir when an error occurs.)

[neonichu]

👍