| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
If the JVM is missing the unlimited cryptography extension, we are
unable to use the AES-192 and AES-256 key sizes. Don't install them
in the server, as the server can't support them if a client requests
it. Instead we log a warning to the server log, so the admin can
troubleshoot the problem if they really require AES-192 or AES-256
for their organization's Gerrit installation.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
If we are setting up an area for the hosted development environment
we probably are just switching from a normal target build to the
weird hosted mode debugging. Rather than blocking on updates for
plugins and so forth, just rely on what is in our repository.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Some installations might be unable to connect to a SMTP relay,
but are still useful through the web page UI, provided that
reviewers check their dashboard periodically. In such cases
we can't open a socket to a SMTP server, so we should bypass
any email sending code paths to prevent errors from filling
up the server logs.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
If the user visits a sidebyside view location, but we come up with
no file hunks to display (e.g. because it really is binary) then
we switch them to the unified viewer, which can show the headers
from the diff.
Bug: GERRIT-193
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
This way test instances don't send spam emails, but also companies
can configure the whitelist to contain only their corporate domain
name and potentially avoid distributing emails outside of their
corporate firewall.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Using a regex pattern of "^(\n|.)*\n..." causes the Java regex package
to recurse for each consideration of the first alternation, creating a
very deep stack for a relatively short commit message. If the message
is large enough, the thread stack overflows, and an exception is thrown.
Instead we use a more customized parser, to examine the message end and
determine if an LF should be appended, or not.
Bug: GERRIT-207
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
| |
Bug: GERRIT-196
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
This removes the dependency on the encumbered Java Mail client made
by Sun Microsystems, and replaces it with a very simple SMTP library
under the Apache License.
Configuration for the email client is now set in gerrit.config, by
the same properties that "git send-email" would honor. Unfortunately
we don't support TLS/SSL as a result of this move.
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>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
The login type, login http header, email format and Google Account
upgrade flags are now specified in gerrit.config.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Now we use user.name, user.email, and auto-detect the committer like
git would normally do.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
This tool can be used to help an admin convert their system settings
data over to the new format.
Currently only sshd_port has been moved to sshd.listenAddress so
that is the only property moved by the tool.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
| |
Otherwise we may not have all of the property values we need.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This was created for Gerrit 2.0.9 to support moving per-project
configuration out of the project's ~/.git/config into the db,
where it was more easily edited by the web UI.
Since the per-project configuration was previously a hidden and
undocumented feature, its likely nobody else used it except for
me on review.source.android.com, and anyone who has used it has
already upgraded past 2.0.9. So this is just dead code now.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
We now support not only the port to listen on, but also the IP address
(or addresses) the daemon should listen on. This may help sites which
host multiple installations on one system, by binding each SSHD onto a
logical IP address.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If the site starts out with a Java formatted "ssh_host_key" file,
but later installs BouncyCastle Crypto, we should continue to use
its existing Java formatted host key, to prevent needing to do a
host key change.
Only one host key (RSA or DSA) is actually required to load the
server and make it functional for clients. Rather than try to
require both keys at server startup we only load ones which are
available on disk.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
If there is a problem writing the generated host key to disk, MINA
logs it at level INFO. We should allow those messages, otherwise
the administrator might not know about a problem.
Bug: GERRIT-210
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
The main change is that PatchScreen now receives the list of all
the file keys in this patch set instead of just the one that is
being reviewed.
Bug: GERRIT-59
|
|
|
|
|
|
|
|
|
|
|
| |
We shouldn't put an entity into the line if the line is empty,
as it shows a space when the user tries to select the line text. In
theory I had put these here to preserve the line height, but that is
already taken care of for us by the line numbers which appear in the
same table row, using the same font.
Bug: GERRIT-211
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Although not yet supported in Ehcache, at least they are
now documented for use in `gerrit flush-caches` calls.
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>
|
|
|
|
|
|
|
|
| |
Since we switched to args4j for our parsing these options are available
on both 'git receive-pack' and 'gerrit receive-pack'. We might as well
just talk about it on 'git receive-pack' in the documentation.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
This item no longer exists in the menus. The Settings link is only
in the right hand area of the menu bar.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
For MySQL in GMT+10 timezones we need to not supply a phony default
value for TIMESTAMP columns in the database. gwtorm 1.1.1 has
removed those defaults.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
We now create a thread pool for each [remote] block found inside of
replication.config. All URLs within that remote block share the same
pool of worker threads. By defining different remote blocks for each
URL the administrator is able to isolate slow WAN links from faster
local warm-spares, ensuring that the more local systems are able to
stay current, even if the WAN gets really far behind.
Bug: GERRIT-200
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
| |
We add and remove frequently, usually on a small number of items,
the set should tend to perform better in terms of avoiding moving
around the other active items in the list.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
We need to use the symbolic ref name when mirroring, as we flatten
it out to a normal ref on the remote site. This prevents cases
where we needed to push both "HEAD" and "refs/heads/master" as the
former was a symbolic-ref to the latter.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
We should (ideally) have a worker thread for each destination
we will replicate too. This permits the thread pool to spawn
each PushOp around on the threads, so the pushes get run in
parallel and don't wind up going sequentially.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
We still use the hypenated form, as that is what JGit is
doing underneath for its defaults.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Now that sync everything tries to use what is effectively ls-remote
to view the remote side refs, we have to do this here.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
| |
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
By supporting substring matches on URLs during "gerrit replicate" an
administrator can reforce replication to a single remote system after
network connectivity with that system is known to be working.
Bug: GERRIT-110
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
|
| |
If replication is enabled, schedule all projects for replication when
the daemon starts. This way if there is any inconsistency between the
local repository and the remotes, it is fixed during startup. We can
find such cases when a delayed push never finished, because the server
was shutdown while it was still sleeping in the pending queue.
Bug: GERRIT-110
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
This way administrators can set the delay for farther away remotes to
be longer, and for really close remotes to be much shorter.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
|
| |
This can be useful to force a project to resync, such as after
detecting a mirror failure and the mirror has been recovered,
or if a ref was deleted from the main project by hand, and it
must be delivered to the mirrors.
Bug: GERRIT-110
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
This way administrative commands can also use this feature to manage
their own database connection, without having to copy the code over.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
This gives the rest of the Gerrit server to come online before we
start hammering the thread pool with a potentially expensive to
perform task (searching the database for outstanding merges, and
spinning up workers to perform them).
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
The task work queue is everything gerrit needs to do in the
background, via some work threads.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
|
|
| |
By extending the ScheduledThreadPoolExecutor we are able to apply a
decoration to the tasks before they are added to the queue, giving
us the ability to pull them out at a later point in time and get any
data we want, in particular, the input Runnable.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|
|
|
|
|
|
|
| |
Its a lot easier to follow what happens here by using small, self
contained methods with clear return types and responsibilities.
Signed-off-by: Shawn O. Pearce <sop@google.com>
|