Bug 34439 – webkit-patch does not prompt username if the default one is incorrect

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Victor Wang]

webkit-patch grabs and shows a default user name when landing a patch. If the default one is incorrect, it should prompt user to change it.
It would be also nice if the script supports a "username" command line option.

[Eric Seidel (no email)]

I'm not sure what the "default" username is.

And is this for bugzilla or svn?

[Daniel Bates]

Not sure if this bug is related, but I can confirm that for svn, unless you have initially checked out the repo. with a username or committed a change by hand (and overridden the username with the --username flag) then the default username is the login username.

(In reply to comment #1)
> I'm not sure what the "default" username is.
> 
> And is this for bugzilla or svn?

[Victor Wang]

This is for svn.

Agree with Daniel, sounds like the default username is from the login username. In this bug, I am more concerning about webkit-patch does not prompt "username" option if the one it tries is incorrect. I would like to be able to specify a username when I run the script or enter the right one if the one the script is using is incorrect.

[Daniel Bates]

Created attachment 48209 [details]
Patch with unit tests

I am open to suggestions on better names for methods and variables as well as suggestions on the approach.

[Adam Barth]

The code around raw_input should be in the user.py module.  Maybe add a retry option to prompt() somehow?  I think we have a similar idiom in bugzilla.py for the bugzilla password that should use the same code.

[Daniel Bates]

Comment on attachment 48209 [details]
Patch with unit tests

Will look into factoring the code based on Adam's comment.

[Daniel Bates]

Created attachment 49015 [details]
Patch with unit tests

[Eric Seidel (no email)]

Comment on attachment 49015 [details]
Patch with unit tests

I would have made this generic:
 307     def has_webkit_authorization_file(home_directory=os.getenv("HOME")):

not webkit-specific.  Or at least have a generic version.  Since it looks at the values from SVN itself, it seems no need to have _webkit_ in the name.  It could use those values as default values to arguments passed to it.

I would have written this using List.extend:
373             svn_commit_args.append('--username')
 374             svn_commit_args.append(username)
svn_commit_args.extend(["--username", username])

I think we're trying to standardize on " instead of ' anyway.  You can ask Adam for more infos.

Making this static just makes it harder to mock/unit test:
6     @staticmethod
 307     def has_webkit_authorization_file(home_directory=os.getenv("HOME")):

I suggest not making it static.

Otherwise I think this looks OK.

Clearly we don't have any unit testing of the actual commit, since the SVN.has_webkit_authorization_file call would actually be hitting disk for the real auth file in those cases!

r- for the above nits.

[Eric Seidel (no email)]

     svn_server_realm = "<http://svn.webkit.org:80> Mac OS Forge"
 261     svn_server_host = "svn.webkit.org"
Should just be instance variables.  They can have some class-level defaults if you like.  But eventually we really need some sort of webkit_config.py module which holds all this webkit-specific stuff.  Not that we ever plan to re-use webkit-patch for some other project, but it makes things read cleaner if we have all the webkit-specifc config in as few places as possible.

[Daniel Bates]

Created attachment 49114 [details]
Patch with unit tests

Updated patch based on Eric's comments.

[Eric Seidel (no email)]

Comment on attachment 49114 [details]
Patch with unit tests

OK.

[WebKit Commit Bot]

Comment on attachment 49114 [details]
Patch with unit tests

Clearing flags on attachment: 49114

Committed r55131: <http://trac.webkit.org/changeset/55131>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.