# ⚠️ Breaking
Many deprecated queue config options are removed (actually, they should
have been removed in 1.18/1.19).
If you see the fatal message when starting Gitea: "Please update your
app.ini to remove deprecated config options", please follow the error
messages to remove these options from your app.ini.
Example:
```
2023/05/06 19:39:22 [E] Removed queue option: `[indexer].ISSUE_INDEXER_QUEUE_TYPE`. Use new options in `[queue.issue_indexer]`
2023/05/06 19:39:22 [E] Removed queue option: `[indexer].UPDATE_BUFFER_LEN`. Use new options in `[queue.issue_indexer]`
2023/05/06 19:39:22 [F] Please update your app.ini to remove deprecated config options
```
Many options in `[queue]` are are dropped, including:
`WRAP_IF_NECESSARY`, `MAX_ATTEMPTS`, `TIMEOUT`, `WORKERS`,
`BLOCK_TIMEOUT`, `BOOST_TIMEOUT`, `BOOST_WORKERS`, they can be removed
from app.ini.
# The problem
The old queue package has some legacy problems:
* complexity: I doubt few people could tell how it works.
* maintainability: Too many channels and mutex/cond are mixed together,
too many different structs/interfaces depends each other.
* stability: due to the complexity & maintainability, sometimes there
are strange bugs and difficult to debug, and some code doesn't have test
(indeed some code is difficult to test because a lot of things are mixed
together).
* general applicability: although it is called "queue", its behavior is
not a well-known queue.
* scalability: it doesn't seem easy to make it work with a cluster
without breaking its behaviors.
It came from some very old code to "avoid breaking", however, its
technical debt is too heavy now. It's a good time to introduce a better
"queue" package.
# The new queue package
It keeps using old config and concept as much as possible.
* It only contains two major kinds of concepts:
* The "base queue": channel, levelqueue, redis
* They have the same abstraction, the same interface, and they are
tested by the same testing code.
* The "WokerPoolQueue", it uses the "base queue" to provide "worker
pool" function, calls the "handler" to process the data in the base
queue.
* The new code doesn't do "PushBack"
* Think about a queue with many workers, the "PushBack" can't guarantee
the order for re-queued unhandled items, so in new code it just does
"normal push"
* The new code doesn't do "pause/resume"
* The "pause/resume" was designed to handle some handler's failure: eg:
document indexer (elasticsearch) is down
* If a queue is paused for long time, either the producers blocks or the
new items are dropped.
* The new code doesn't do such "pause/resume" trick, it's not a common
queue's behavior and it doesn't help much.
* If there are unhandled items, the "push" function just blocks for a
few seconds and then re-queue them and retry.
* The new code doesn't do "worker booster"
* Gitea's queue's handlers are light functions, the cost is only the
go-routine, so it doesn't make sense to "boost" them.
* The new code only use "max worker number" to limit the concurrent
workers.
* The new "Push" never blocks forever
* Instead of creating more and more blocking goroutines, return an error
is more friendly to the server and to the end user.
There are more details in code comments: eg: the "Flush" problem, the
strange "code.index" hanging problem, the "immediate" queue problem.
Almost ready for review.
TODO:
* [x] add some necessary comments during review
* [x] add some more tests if necessary
* [x] update documents and config options
* [x] test max worker / active worker
* [x] re-run the CI tasks to see whether any test is flaky
* [x] improve the `handleOldLengthConfiguration` to provide more
friendly messages
* [x] fine tune default config values (eg: length?)
## Code coverage:
# Why this PR comes
At first, I'd like to help users like #23636 (there are a lot)
The unclear "Internal Server Error" is quite anonying, scare users,
frustrate contributors, nobody knows what happens.
So, it's always good to provide meaningful messages to end users (of
course, do not leak sensitive information).
When I started working on the "response message to end users", I found
that the related code has a lot of technical debt. A lot of copy&paste
code, unclear fields and usages.
So I think it's good to make everything clear.
# Tech Backgrounds
Gitea has many sub-commands, some are used by admins, some are used by
SSH servers or Git Hooks. Many sub-commands use "internal API" to
communicate with Gitea web server.
Before, Gitea server always use `StatusCode + Json "err" field` to
return messages.
* The CLI sub-commands: they expect to show all error related messages
to site admin
* The Serv/Hook sub-commands (for git clients): they could only show
safe messages to end users, the error log could only be recorded by
"SSHLog" to Gitea web server.
In the old design, it assumes that:
* If the StatusCode is 500 (in some functions), then the "err" field is
error log, shouldn't be exposed to git client.
* If the StatusCode is 40x, then the "err" field could be exposed. And
some functions always read the "err" no matter what the StatusCode is.
The old code is not strict, and it's difficult to distinguish the
messages clearly and then output them correctly.
# This PR
To help to remove duplicate code and make everything clear, this PR
introduces `ResponseExtra` and `requestJSONResp`.
* `ResponseExtra` is a struct which contains "extra" information of a
internal API response, including StatusCode, UserMsg, Error
* `requestJSONResp` is a generic function which can be used for all
cases to help to simplify the calls.
* Remove all `map["err"]`, always use `private.Response{Err}` to
construct error messages.
* User messages and error messages are separated clearly, the `fail` and
`handleCliResponseExtra` will output correct messages.
* Replace all `Internal Server Error` messages with meaningful (still
safe) messages.
This PR saves more than 300 lines, while makes the git client messages
more clear.
Many gitea-serv/git-hook related essential functions are covered by
tests.
---------
Co-authored-by: delvh <dev.lh@web.de>
Change all license headers to comply with REUSE specification.
Fix#16132
Co-authored-by: flynnnnnnnnnn <flynnnnnnnnnn@github>
Co-authored-by: John Olheiser <john.olheiser@gmail.com>
* clean git support for ver < 2.0
* fine tune tests for markup (which requires git module)
* remove unnecessary comments
* try to fix tests
* try test again
* use const for GitVersionRequired instead of var
* try to fix integration test
* Refactor CheckAttributeReader to make a *git.Repository version
* update document for commit signing with Gitea's internal gitconfig
* update document for commit signing with Gitea's internal gitconfig
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Follows #19266, #8553, Close#18553, now there are only three `Run..(&RunOpts{})` functions.
* before: `stdout, err := RunInDir(path)`
* now: `stdout, _, err := RunStdString(&git.RunOpts{Dir:path})`
This PR registers requests with the process manager and manages hierarchy within the processes.
Git repos are then associated with a context, (usually the request's context) - with sub commands using this context as their base context.
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Second attempt at preventing zombies
* Ensure that the pipes are closed in ssh.go
* Ensure that a cancellable context is passed up in cmd/* http requests
* Make cmd.fail return properly so defers are obeyed
* Ensure that something is sent to stdout in case of blocks here
Signed-off-by: Andrew Thornton <art27@cantab.net>
* placate lint
Signed-off-by: Andrew Thornton <art27@cantab.net>
* placate lint 2
Signed-off-by: Andrew Thornton <art27@cantab.net>
* placate lint 3
Signed-off-by: Andrew Thornton <art27@cantab.net>
* fixup
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Apply suggestions from code review
Co-authored-by: 6543 <6543@obermui.de>
Co-authored-by: Lauris BH <lauris@nix.lv>
* _ to unused func options
* rm useless brakets
* rm trifial non used models functions
* rm dead code
* rm dead global vars
* fix routers/api/v1/repo/issue.go
* dont overload import module
There is a potential memory leak in `Workerpool` due to the intricacies of
`time.Timer` stopping.
Whenever a `time.Timer` is `Stop`ped its channel must be cleared using a
`select` if the result of the `Stop()` is `false`.
Unfortunately in `Workerpool` these were checked the wrong way round.
However, there were a few other places that were not being checked.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
* Delay printing hook statuses until after 1 second
* Move to a 5s delay, wrapped writer structure and add config
* Update cmd/hook.go
* Apply suggestions from code review
* Update cmd/hook.go
Co-authored-by: Antoine GIRARD <sapk@users.noreply.github.com>
#8982 attempted to enforce the gitea environment for pushes - unfortunately it tested the settings before they were actually read in - and therefore does not do that!
* make notifyWatchers work on multiple actions
* more efficient multiple notifyWatchers
* Make CommitRepoAction take advantage of multiple actions
* Batch post and pre-receive results
* Set batch to 30
* Auto adjust timeout & add logging
* adjust processing message
* Add some messages to pre-receive
* Make any non-200 status code from pre-receive an error
* Add missing hookPrintResults
* Remove shortcut for single action
* mistaken merge fix
* oops
* Move master branch to the front
* If repo was empty and the master branch is pushed ensure that that is set as the default branch
* fixup
* fixup
* Missed HookOptions in setdefaultbranch
* Batch PushUpdateAddTag and PushUpdateDelTag
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Add an option to protected branches to add writing deploy keys to the whitelist for pushing.
Please note this is technically a breaking change: previously if the owner of a repository was on the whitelist then any writing deploy key was effectively on the whitelist. This option will now need to be set if that is desired.
Closes#8472
Details:
* Allow Protected Branches to Whitelist Deploy Keys
* Add migration
* Ensure that IsDeployKey is set to false on the http pushes
* add not null default false
* Fix#6946 by checking PullRequest ID on pushing
* Ensure we have the owner name, the pr attributes and the the issue
* Fix TestSearchRepo by waiting till indexing is done
* Update integrations/repo_search_test.go
* changes as per @mrsdizzie
* missing comma
* Spelling mistake
* Fix full pushing environment
* Move hook functionality internally
* Internalise serv logic
* Remove old internal paths
* finally remove the gitlogger
* Disallow push on archived repositories
* fix lint error
* Update modules/private/key.go
* Update routers/private/hook.go
* Update routers/private/hook.go
* Update routers/private/hook.go
* Updated routers/private/serv.go
* Fix LFS Locks over SSH
* rev-list needs to be run by the hook process
* fixup
* Improve git test
* Ensure that the lfs files are created with a different prefix
* Reduce the replication in git_test.go
* slight refactor
* Remove unnecessary "/"
* Restore ensureAnonymousClone
* Restore ensureAnonymousClone
* Run rev-list on server side
* Try passing in the alternative directories instead
* Mark test as skipped
* Improve git test
* Ensure that the lfs files are created with a different prefix
* Reduce the replication in git_test.go
* Remove unnecessary "/"
When replicating to gitea from a remote system which makes use of
git refs to store extra data (for example, gerrit), pushing a lot
of refs to gitea can cause problems due to the extra processing
that the pre and post receive hooks perform. But it's still
useful for gitea to be able to serve those refs. This change
skips unecessary processing of refs other than branches or tags.
We don't need to check any ref that isn't a branch for branch
protection (protection will never be enabled). So in the
pre-receive hook, we wrap that check in a test for whether the
ref is a branch.
We also don't need to add information to the activity stream about
pushes to non-standard refs, so we skip that step in the
post-receive hook for refs which are not branches or tags.
For some concrete examples, gerrit maintains a ref for every
patchset of every change in the form refs/changes/XX/YYYY/Z.
Many systems use refs/notes to store additonal data about commits.
This change allows these and other schemes to be used without
affecting gitea.
* Panic don't fatal on create new logger
Fixes#5854
Signed-off-by: Andrew Thornton <art27@cantab.net>
* partial broken
* Update the logging infrastrcture
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Reset the skip levels for Fatal and Error
Signed-off-by: Andrew Thornton <art27@cantab.net>
* broken ncsa
* More log.Error fixes
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Remove nal
* set log-levels to lowercase
* Make console_test test all levels
* switch to lowercased levels
* OK now working
* Fix vetting issues
* Fix lint
* Fix tests
* change default logging to match current gitea
* Improve log testing
Signed-off-by: Andrew Thornton <art27@cantab.net>
* reset error skip levels to 0
* Update documentation and access logger configuration
* Redirect the router log back to gitea if redirect macaron log but also allow setting the log level - i.e. TRACE
* Fix broken level caching
* Refactor the router log
* Add Router logger
* Add colorizing options
* Adjust router colors
* Only create logger if they will be used
* update app.ini.sample
* rename Attribute ColorAttribute
* Change from white to green for function
* Set fatal/error levels
* Restore initial trace logger
* Fix Trace arguments in modules/auth/auth.go
* Properly handle XORMLogger
* Improve admin/config page
* fix fmt
* Add auto-compression of old logs
* Update error log levels
* Remove the unnecessary skip argument from Error, Fatal and Critical
* Add stacktrace support
* Fix tests
* Remove x/sync from vendors?
* Add stderr option to console logger
* Use filepath.ToSlash to protect against Windows in tests
* Remove prefixed underscores from names in colors.go
* Remove not implemented database logger
This was removed from Gogs on 4 Mar 2016 but left in the configuration
since then.
* Ensure that log paths are relative to ROOT_PATH
* use path.Join
* rename jsonConfig to logConfig
* Rename "config" to "jsonConfig" to make it clearer
* Requested changes
* Requested changes: XormLogger
* Try to color the windows terminal
If successful default to colorizing the console logs
* fixup
* Colorize initially too
* update vendor
* Colorize logs on default and remove if this is not a colorizing logger
* Fix documentation
* fix test
* Use go-isatty to detect if on windows we are on msys or cygwin
* Fix spelling mistake
* Add missing vendors
* More changes
* Rationalise the ANSI writer protection
* Adjust colors on advice from @0x5c
* Make Flags a comma separated list
* Move to use the windows constant for ENABLE_VIRTUAL_TERMINAL_PROCESSING
* Ensure matching is done on the non-colored message - to simpify EXPRESSION
* Use PathUnescape instead of QueryUnescape when working with branch names
Currently branch names with a '+' fail in certain situations because
QueryUnescape replaces the + character with a blank space.
Using PathUnescape should be better since it is defined as:
// PathUnescape is identical to QueryUnescape except that it does not
// unescape '+' to ' ' (space).
Fixes#6333
* Change error to match new function name
* Add new util function PathEscapeSegments
This function simply runs PathEscape on each segment of a path without
touching the forward slash itself. We want to use this instead of
PathEscape/QueryEscape in most cases because a forward slash is a valid name for a
branch etc... and we don't want that escaped in a URL.
Putting this in new file url.go and also moving a couple similar
functions into that file as well.
* Use EscapePathSegments where appropriate
Replace various uses of EscapePath/EscapeQuery with new
EscapePathSegments. Also remove uncessary uses of various
escape/unescape functions when the text had already been escaped or was
not escaped.
* Reformat comment to make drone build happy
* Remove no longer used url library
* Requested code changes
* move all database operations from hook command to web command and instead of internal routes
* bug fixed
* adjust the import path sequences
* remove unused return value on hookSetup
wxiaoguang
