Backport #20925
This commit updates the `GET /api/v1/repos/{owner}/{repo}/archive/{archive}`
endpoint which prior to this PR had a couple of issues.
1. The endpoint had a hard-coded 20s timeout for the archiver to complete after
which a 500 (Internal Server Error) was returned to client. For a scripted
API client there was no clear way of telling that the operation timed out and
that it should retry.
2. Whenever the timeout _did occur_, the code used to panic. This was caused by
the API endpoint "delegating" to the same call path as the web, which uses a
slightly different way of reporting errors (HTML rather than JSON for
example).
More specifically, `api/v1/repo/file.go#GetArchive` just called through to
`web/repo/repo.go#Download`, which expects the `Context` to have a `Render`
field set, but which is `nil` for API calls. Hence, a `nil` pointer error.
The code addresses (1) by dropping the hard-coded timeout. Instead, any
timeout/cancelation on the incoming `Context` is used.
The code addresses (2) by updating the API endpoint to use a separate call path
for the API-triggered archive download. This avoids producing HTML-errors on
errors (it now produces JSON errors).
Signed-off-by: Peter Gardfjäll <peter.gardfjall.work@gmail.com>
Signed-off-by: Peter Gardfjäll <peter.gardfjall.work@gmail.com>
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Peter Gardfjäll <peter.gardfjall.work@gmail.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Backport #20902
When setting.Git.DisablePartialClone is set to false then the web server will add filter support to web http. It does this by using`-c` command arguments but this will not work on gitea serv as the upload-pack and receive-pack commands do not support this.
Instead we move these options into the .gitconfig instead.
Fix #20400
Signed-off-by: Andrew Thornton <art27@cantab.net>
Signed-off-by: Andrew Thornton <art27@cantab.net>
Backport #20967
Currently, it's impossible to connect to self-signed TLS encrypted redis instances. The problem lies in inproper error handling, when building redis tls options - only invalid booleans are allowed to be used in `tlsConfig` builder. The problem is, when `strconv.ParseBool(...)` returns error, it always defaults to false - meaning it's impossible to set `tlsOptions.InsecureSkipVerify` to true.
Fixes #19213
Co-authored-by: Igor Rzegocki <ajgon@users.noreply.github.com>
Backport #21011
When migrating add several more important sanity checks:
* SHAs must be SHAs
* Refs must be valid Refs
* URLs must be reasonable
Signed-off-by: Andrew Thornton <art27@cantab.net>
This should solve the main problem of dynamic assets getting stale after
a version upgrade. Everything not affected will use query-string based
cache busting, which includes files loaded via HTML or worker scripts.
Backport #20847
If permissions are incorrect for writing to the doctor log simply disable the log file
instead of panicing.
Related #20570
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: delvh <dev.lh@web.de>
Backport #20893
In the compare endpoint the git fetch is restricted to a certain branch however,
this does not completely prevent tag acquisition/pollution as git fetch will collect
any tags on that branch.
This causes pollution of the tag namespace and could cause confusion by users.
This PR adds `--no-tags` to the `git fetch` call.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Don't open new page for ext wiki on same repository (#20725)
- Backport of #20725
- When the external wiki has been set to a file on the repository, don't open the page on a tab.
- Resolves #20657
* Gofmt
* Fix line
Co-authored-by: zeripath <art27@cantab.net>
The graceful manager waits for 4 listeners to be created or to be told that they are not needed. If it is not told about them it will indefinitely and timeout.
This leads to SVC hosts not being told of being in the readyState but on Unix would lead to the termination of the process.
There was an unfortunate regression in #20299 which missed this subtly and in the case whereby SSH is disabled the `builtinUnused()` is not called.
This PR adds a call to `builtinUnused()` when not using the builtin ssh to allow `createServerWaitGroup.Done()` to be called.
In addition it was noted that the if/else clauses for timeout informing of the SVC host were in the wrong order. These have been swapped.
Fix #20609
Backport #20637
Windows doesn't have the concept of "executable" POSIX bits so for now always return true to minimise doctor and logging noise. Addresses #20636
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: JonRB <4564448+eeyrjmr@users.noreply.github.com>
Co-authored-by: silverwind <me@silverwind.io>
* Add disable download source configuration (#20548)
Add configuration to enable/disable download source from UI.
Co-authored-by: zeripath <art27@cantab.net>
* Fix BaseVars not used in renderering
* Fix disabled open in vscode menu when disabling download source from UI
Co-authored-by: zeripath <art27@cantab.net>
Co-authored-by: Lauris BH <lauris@nix.lv>
This enables git.Command's Run to optionally use the given context directly so its deadline will be respected. Otherwise, it falls back to the previous behavior of using the supplied timeout or a default timeout value of 360 seconds.
repo's serviceRPC() calls now use the context's deadline (which is unset/unlimited) instead of the default 6-minute timeout. This means that large repo clones will no longer arbitrarily time out on the upload-pack step, and pushes can take longer than 6 minutes on the receive-pack step.
Fixes #20680
* Add username check to doctor
- Add a new breaking change detector to Gitea's doctor, which checks if
all users still have a valid username according to Gitea. Given from
time-to-time we need to make changes, either due to new routes or due to
security, it's for a instance's admin to check if all users still have a
valid username.
* Fix extra argument
* Apply suggestions from code review
* Apply suggestions from code review
The code for detection of lines in highlight.go is somewhat too complex
and doesn't take account of how Chroma is actually splitting things into
lines for us.
Remove both the .line and .cl classes from Chroma's HTML which made
the old conditional work again. This fixed Copy of YAML files while also
reducing the amount of rendered HTML nodes.
Co-authored-by: zeripath <art27@cantab.net>
Co-authored-by: Lauris BH <lauris@nix.lv>
- Backport of #20616
- Git only decides to use the Wire 2 protocol when `git {receive,upload}-pack` receive the `GIT_PROTOCOL` environment with as value `version=2`. Currently the internal SSH Server wasn't passing this environment through. The `gitea serv` code already passed all received environments to the git command, so no code changes there.
- This is mentioned in Git manual, https://git-scm.com/docs/git#Documentation/git.txt-codeGITPROTOCOLcode
- Always respect the user's configured mime type map
- Allow more types like image/pdf/video/audio to serve with correct content-type
- Shorten cache duration of raw files to 5 minutes, matching GitHub
- Don't set `content-disposition: attachment`, let the browser decide whether it wants to download or display a file directly
- Implement rfc5987 for filenames, remove previous hack. Confirmed it working in Safari.
- Make PDF attachment work in Safari by removing `sandbox` attribute.
This change will make a lot more file types open directly in browser now. Logic should generally be more readable than before with less `if` nesting and such.
Replaces: https://github.com/go-gitea/gitea/pull/20460
Replaces: https://github.com/go-gitea/gitea/pull/20455
Fixes: https://github.com/go-gitea/gitea/issues/20404
Backport #20476
The code in modules/ssh/ssh.go:sessionHandler() currently cause an error to be
logged if `gitea serv` exits with a exit(1). This logging is useless because the
accompanying stderr is not provided and in any case the exit(1) is most likely due
to permissions errors.
Further it then causes the EOF to be logged - even though this is not helpful.
This PR simply checks the errors returned and stops logging them.
In the case of misconfigurations causing `gitea serv` to fail with exit(1)
the current logging is not helpful at determining this and users should simply
review the message passed over the ssh connection.
Fix #20473
Signed-off-by: Andrew Thornton <art27@cantab.net>
Backport #20412
A lot of existing packages do not conform to SemVer, yet, they should be allowed
in the Conan package registry as-is. To achieve this, remove the SemVer check
from `NewRecipeReference`, and replace it with a simple empty string check.
A unit test with a non-semver version is also included.
Fixes #20405.
Signed-off-by: Gergely Nagy <me@gergo.csillger.hu>
Co-authored-by: KN4CK3R <admin@oldschoolhack.me>
Co-authored-by: Gergely Nagy <algernon@users.noreply.github.com>
`no-transform` allegedly disables CloudFlare auto-minify and we did not
set caching headers on html or api requests, which seems good to have
regardless.
Transformation is still allowed for asset requests.
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Backport #20272
Unfortunately it appears that 2048 bit RSA keys can occasionally be created in such
a way that they appear to have 2047 bit length. This PR simply changes our defaults to
allow these.
Fix #20249
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
When viewing a subdirectory and the latest commit to that directory in
the table, the commit status icon incorrectly showed the status of the
HEAD commit instead of the latest for that directory.
Backport #20346
Although `WalkGitLog` tries to test for `context.DeadlineExceededErr`
there is a small chance that the error will propagate to the reader
before it is recognised. This will cause the error to propagate up to
`renderDirectoryFiles` and cause a http status 500.
Here we check that the error passed is a `DeadlineExceededErr` via error.Is
Fix #20329
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: zeripath <art27@cantab.net>
Backport #20299. Follow #20298. Only the `GlobalInitInstalled` function should prepare the SSH files for external server or starts the builtin server.
* `trustedUserCaKeys` is removed, use `SSH.TrustedUserCAKeys` directly
* introduce `ssh.Init`, move the SSH init code from `routers/init.go` to it
* `ssh.Init` will start builtin SSH server or prepare external SSH server files
Backport #20289
The code is as old as back to 2016, creating the directory automatically is not correct IMO.
In other places for ssh key writing (RewriteAllPrincipalKeys / appendAuthorizedKeysToFile, etc), the directory will still be created when updating the keys.
This PR will resolve the confusing and annoying problem: the dummy and empty ".ssh" directory in new git home
Before, in #19732, the old home directory is not correct.
This PR introduces a new config option for git home: git.HOME_PATH,
which is default to %(APP_DATA_PATH)/home
And pass env GNUPGHOME to git command, force Gitea to use a stable GNUPGHOME directory
- Backport (#20277
- This code is only valid when `refNumeric` exist(otherwise we didn't find such numeric PR and can skip that check) and give a free-pas to the "BEFORE" check when `ref` is nil.
- Resolves #20109
* Disable federation by default (#20045)
- Backport #20045
- A Gitea instance should choose whetever they want to federate(as once it has more features also brings extra costs/moderation/unexpected behavior) with other AP/ForgeFed software.
* Fix tests
Fixes #19090
If the user-agent starts with git and user must change password but
hasn't return a 401 with the message.
It must be a 401, git doesn't seem to show the contents of the error message
when we return a 403
Co-authored-by: 6543 <6543@obermui.de>
It appears possible that stopwatches can become orphaned or have been orphaned in
the past.
This PR adds Orphan checks for Stopwatches.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
* Empty log queue on flush and close
It is possible for log events to remain in the buffer off the multichannelledlog
and thus not be logged despite close or flush.
This PR simply adds a function to empty the queue before closing or flushing.
(Except when the logger is paused.)
Reference #19982
Signed-off-by: Andrew Thornton <art27@cantab.net>
* and do similar for ChannelledLog
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
If there are dangling stopwatches with missing issues there will be repeated
logging of Unable to APIFormat stopwatches. These are unhelpful and instead
we should only log if the error is not an issue not exist error.
And we should also prevent an error on missing issue in GetActiveStopwatch too
Signed-off-by: Andrew Thornton <art27@cantab.net>
