| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
This brings in bug fixes related to Key object's fromString not
decoding enough steps before returning the data to the application.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
If the user modifies their preferred email address, enable
the save button so that they can actually save the change.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
Change tables now display a hyperlink to view all open changes for
that project, browsing them in the same order that the all open
changes for the entire site would navigate them in.
Bug: GERRIT-32
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This style of link is better suited to copy and pasting into a longer
term repository, like a bug tracking system. I keep finding myself
copying change ids from Gerrit into Gerrit's own issue tracker, but
I edit them by hand to be permalink style. Using that in the output
just makes it easier on me and everyone else.
I highly doubt the speed improvement of using the anchor form is a
major issue for us, users copying links from a terminal are likely
going to be opening a new browser tab/window for the link rather
than recycling one that already has Gerrit open.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
We don't need these errors in our application server log. If the
client has given us a hostname we can't resolve, trap the error
and return an invalid discovery JSON payload to them instead. The
user will see a message explaining its an unsupported/invalid URL
and can try again.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Its horrible to construct an editor into the line number column;
it completely throws off the UI layout of the entire window.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Otherwise the dashboard is empty, and it looks stupid.
Bug: GERRIT-49
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Some users may want to automatically hide the site header/footer
assets upon sign-in, as they might take up a lot of screen space.
For example, the site header at http://review.source.android.com/
is 86 pixels tall.
Bug: GERRIT-45
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
The server administrator really doesn't care about the text
we are showing to the end-user to let them know the CLA is
missing or expired.
Bug: GERRIT-53
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Messages like:
2009-01-28 13:21:56,884::DEBUG: org.apache.sshd.common.channel.Window - Increase server local window by 262263 up to 2088960
aren't that interesting to the server administrator.
Bug: GERRIT-52
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change shows "git pull ..." command lines for fetching
a patch set down off Gerrit using standard Git tools, rather
than relying upon the "repo download" command. It also makes
the usage of "repo download" more optional by not showing the
repo download command line by default.
If a project is not readable by anonymous users then we offer
an SSH URL instead of an anonymous git:// URL.
Bug: GERRIT-55
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
Sorting by the quite stable change id is reasonable. On current
Gerrit2 systems its sequential, due to having a single sequence
generator assigning change ids in order.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
This may help automation tools to know the change id and the git
commit associated with the message, without needing to run off to
the Gerrit server's JSON interface, or rip apart the X-Change-URL
header to obtain the id component from it.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
GMail (and probably a lot of other email clients) break a thread
if the List-ID differs between the original message and any type
of reply to that original message. This causes comments made on
a change to come up in a different thread than the original notice
message, which makes it harder to organize.
We now put the message type into an X-Gerrit-MessageType header,
in case a client wants to filter on it, and thus way we can keep
the List-ID constant throughout the "thread".
Bug: GERRIT-57
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
The entire point of the code review systems is to have another
party beyond the uploader review the changes to determine they
are suitable for inclusion, and set the approval bits onto the
change record. Automatically approving your own changes reduces
the value of this 3rd party review.
Bug: GERRIT-56
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
This way its formatted as a paragraph in both email clients and
the web UI, which makes it look a lot nicer.
Bug: GERRIT-58
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
Because the text is actually inside a gwt Hyperlink object its down
inside of a div, not the td directly. We need to apply the font
controls onto the div, otherwise the div overrides them and we don't
get the change subject to truncate nicely when its too long for the
available screen space.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
This makes it easier to store multiple editions of the documentation
for Gerrit 2.0, 2.1, and so on. Google AppEngine has a 1,000 file
limit, so storing more than a few editions online would be difficult.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This command evaluates through all known projects and outputs their
name, one per line, if the current user account has Read access to
that project. Its mostly useful for scripting applications such as
mirroring what you can see:
for p in $(ssh -p 29418 host gerrit ls-projects)
do
git clone git://mirror/$p.git
done
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
This way we can add more commands to the SSH library and let them
access the account group data for the current user, as well as test
the user's access rights against projects.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
The process-classes phase is too late in the lifecycle, Maven may
have already embedded our old properties file with some previous
version number in it.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
If we do, we get null pointer exceptions while trying
to evaluate the rules.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Four new rights can be used to control access over the SSH server:
* Push Annotated Tag (refs/tags/*)
* Push Branch: (refs/heads/*)
- Update Existing
- Create New
- Force Update (aka non-fast forward) ; Delete
These can be granted on a per-group/per-project basis to permit users
to remotely control branches over the Git SSH interface. Its mostly
useful for groups that are used to having a standard Git server, where
"anything goes" is permitted within the scope of the project.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
I missed the ", txn" on each of these insert calls, allowing the
category to possibly create without its range of values.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
This paves the way for Git based SSH commands operating on a
project to test other types of approval categories beyond the
basic "READ" category used for coarse-grained access controls.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
If a ref has been removed from the local repository but is in the
push queue, the caller wants to remove it from the local repository
and also from the replicated remotes. Rather than crash out with
an IOException we now push the delete request.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Earlier I added UserAgent.gwt.xml but failed to ensure it
would show up in the JAR produced by Maven.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
Trying to use the <iframe> based sign-in on an Android or
an iPhone/iPod mobile device is very difficult. The total
screen space is just too small to float the window onto.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Users who login with a delegate OpenID want to record their
claimed identity, not their delegate identity.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
Google Analytics won't track the anchor of a page, so we need to
do a search-replace for '#' with '/', which is something that it
will track. Thus tracked URLs would be "/Gerrit/change,1".
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
| |
This is automatically imported by the stock GWT User module.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Excessively long lines force the file to be very hard to read in
a side-by-side presentation as it makes one (or both) columns
too wide for many common display devices, especially laptops.
Forcing our own line wrapping at some reasonable limit (e.g. 100)
ensures that even if the project's style standards permit very long
lines the page is still at least readable on a smaller monitor.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Suggestions were failing because PostgreSQL was treating \uffff
as smaller than 'a', so we never selected the range of possible
candidates and instead always wound up with an empty result set,
unless we had the input string exactly equal to some existing
record in the query.
This max character of \u9fa5 was experimentally determined by
loading all characters between \u0001 and \uffff into a single
PostgreSQL table, then selecting the one record returned by
the query "SELECT x FROM t ORDER BY x DESC LIMIT 1".
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
This is slightly better than showing the name as "undefined", due
to the full name field being Java null/JavaScript undefined.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
Redirect URLs for /r/$commitid were not working correctly if the
collate setting for the database is en_US.UTF-8. Using '~' does
not produce the correct results as some hex digits may sort after
the '~'. Instead use 'z' as this would never be a valid hex digit.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
We should only include the "--\n" line in an email sent from Gerrit
if there will be content below that line. Otherwise its just noise
and looks out of place in the message.
Usually there is content below the line, as the canonical_url is set
and thus we can make backlinks to ourselves. But this isn't always
the case, so we should format a bit better when it isn't.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
If we don't have a URL with which to render links back to Gerrit
(such as if system_config isn't set with a canonical_url) then
we cannot append the settings URL, as it is "null".
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
This way if a new reviewer was added by the "repo upload" tool they
get notification of the change in their inbox.
This also gives existing reviewers a "ping" to let them know that the
uploader wants them to look at the change again.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
If a change is updated with a new patch set the caller may have given
us additional reviewers to include on that change. Those should be
added to the change just like if the change was new, so that those
individuals can see the change in their dashboard.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Aside from emailing a reviewer with a change notice we also want
to ensure the change is visible in their dashboard, should they
just visit the site directly without following a hyperlink.
To do this we need to ensure at least one ChangeApproval has been
created for the reviewer for this change.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
Repeated uses of "repo upload --replace" for the same change id
and commit, without changing the commit SHA-1, would create more
than one patch set with identical content. This is confusing to
read and review. Disallowing it simplifies things for any reader.
Caused-by: jbq
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The "repo upload" client always sends us a refs/for/$branch command,
even if the client is only trying to add a new patch set to an
already existing change entity. It does so because this simplifies
the client logic from needing to handle the case of "replacing
everything" vs. "replacing some, creating some".
We shouldn't consider it an error if the client didn't ask us to make
anything new, but did supply at least one patch set replace to us.
Noticed-by: jbq
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
We covered this in the Eclipse specific setup, but not in the
general setup case. Either way you need to make sure these
packages are installed into your local Maven repository.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|