In Python 2.7 xmlrpclib was changed to use httplib.HTTP[S]Connection instead of the deprecatede httplib.HTTP[S]. This broke our xmlrpclib.Transport subclass which overrode make_connection(). For compatibility with old and new Python, switch to subclassing Transport and SafeTransport separately, with a mixin to add cookies. https://bugzilla.gnome.org/show_bug.cgi?id=628731
…onally using commit.subject By mistake, the code was discarding the description after parsing it out from the edit buffer. https://bugzilla.gnome.org/show_bug.cgi?id=628455
The example url in the section "Ways to refer to a bug" was causing asciidoc to emit invalid xml (by closing a docbook ulink tag before the closing semicolon for a > replacement giving ></ulink>;) which meant that xmlto could not then produce a man page. Quoting the url seems to workaround this issue. asciidoc version was 8.4.4 https://bugzilla.gnome.org/show_bug.cgi?id=621580
Rather than taking empty or unknown input as a "no", just reprompt. I don't think the default value is obvious enough for people to actually figure it out and rely on it, so we might as well add extra robustness to accidental input.
Python adds a space character between consecutive calls to print when the first print doesn't end in a newline. To avoid problems with this, use <file>.write() rather than print when: * We might be printing something empty * The newline after the line is added via echoed user input on the terminal Tracked down by William Jon McCann <firstname.lastname@example.org>
Chromium and Google Chrome have the same basic way of storing cookes, but different config paths. Add a google-chrome browser option with the correct config path for Google Chrome. Based on a patch by Eitan Isaacson <email@example.com>
When they moved to webkit as the browser engine, epiphany changed the location of their cookies database, but the format remained the same. This patch adds support for the new location, but falls back to the old location if the new one doesn't exist. https://bugzilla.gnome.org/show_bug.cgi?id=609668
The code that was supposed to handle 'See Mozilla bug http://bugzilla.mozilla.org/show_bug.cgi?...'; as a cross reference was also detecting 'See https://developer.mozilla.org/en/SpiderMonkey/JSAPI_Reference/JS_MaybeGC https://bugzilla.gnome.org/show_bug.cgi?id=614725'; As a cross-reference. Fix by requiring the "intervening words" not to have a : or / in them, since "See <url>" is clearly standalone and something with a : or a / in it is probably an URL. Reported by Adel Gadllah https://bugzilla.gnome.org/show_bug.cgi?id=619328
When we check for uncommitted changes prior to adding URLs to to commit messages we should ignore submodules, since they don't affect the rebase xprocess we use to add URLs. Reported by Benjamin Otte https://bugzilla.gnome.org/show_bug.cgi?id=611693
I had accidentally prepended the bugzilla installation path in both BugServer.send_post() and BugServer.send_request() (which is called by send_post()), so the path was getting prepended twice when trying to attach a patch.
In BugHandle.parse(), we first try to parse the bug reference as a url and then fall back to interpreting it as an alias from the config file. Unfortunately, something like "gnome:123456" gets parsed as a url (with a scheme of 'gnome' and a path of '123456', and no hostname). This resulted in us passing a None hostname to tracker_get_auth_user(), which resulted in an uncaught exception. From now on, only proceed with treating the bug reference as a url if the scheme is parsed as 'http' or 'https'.
This patch should support http authentication credentials specified in url form (e.g. https://user:firstname.lastname@example.org) or in a configuration file. The additional configure options are: - bz-tracker.<alias>.authuser - bz-tracker.<alias>.authpwd
Currently, git-bz assumes that all bugzilla installations are at the root of the host (e.g. http://foo.com/). It is currently impossible to use git-bz with a bugzilla installation that is hosted at e.g. foo.com/bugzilla/. This patch adds that ability. A new configuration option ('bz-tracker.alias.path') allows the path to be specified for a host.
Handle redirect HTTP responses, in particular if a Bugzilla server is redirecting from http to https. We try to detect "Bugzilla URL base is over here" when we ask for show_bug.cgi and remember that for future requests to the same BugServer to avoid too many redirections. Switch from caching connection on the BugServer to a global connection cache, and rewrite the BugServer cache so to deal with the possibility of redirections.
I'm not entirely up to speed with the new syntax of Python 2.6, but apparently you can now do: func(*args, named_arg=value, **options) and: func(*args, named_arg=value) and Python will happily coalesce named_arg into **options for you. This is not a valid syntax for Python 2.5, though.
Instead of resetting the branch and applying the patches there, move to a detached HEAD, apply patches to the detached HEAD and then make a single update to the branch ref. The only real advantage of this is that 'git reflog show <branchname>' shows a single meaningful commit.