Gitea runs diff on highlighted code fragment for each line in order to provide
code highlight diffs. Unfortunately this diff algorithm is not aware that span tags
and entities are atomic and cannot be split.
The current fixup code makes some attempt to fix these broken tags however, it cannot
handle situations where a tag is split over multiple blocks.
This PR provides a more algorithmic fixup mechanism whereby spans and entities are
completely coalesced into their respective blocks.
This may result in a incompletely reduced diff but - it will definitely prevent the
broken entities and spans that are currently possible.
As a result of this fixup several inconsistencies were discovered in our testcases
and these were also fixed.
Fix #14231
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Handle incomplete diff files properly
The code for parsing diff hunks has a bug whereby a very long line
in a very long diff would not be completely read leading to an unexpected
character.
This PR ensures that the line is completely cleared
Fix #13602
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Also allow git max line length <4096
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Add test case
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
* Alternative fix for HTML diff entity split
This commit both reverts PR #13357 and uses the exiting implementation alredy used for spans to fix the same issue. That PR duplicates most of logic that is already present elsewhere and still was failing for some cases. This should be simpler as it uses the existing logic that already works for <span>s being split apart.
Added both test cases as well.
* Update gitdiff_test.go
* fmt
* entity can have uppercase letter, also add detailed comment per @zeripath
* When creating line diffs do not split within an html entity
Fix #13342
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Add test case
Signed-off-by: Andrew Thornton <art27@cantab.net>
* improve test
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
* Fix initial commit page
Unfortunately as a result of properly fixing ParsePatch the hack that
used git show <initial_commit_id> to get the diff for this failed.
This PR fixes this using the "super-secret" empty tree ref to make the
diff against.
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Also fix #13248
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Update services/gitdiff/gitdiff.go
Co-authored-by: 6543 <6543@obermui.de>
* Add better error checking for inline html diff code
A better fix for #13191 which cleans up this code a bit and adds basic checking which should avoid writing broken HTML in future situations.
* Update gitdiff_test.go
* better regex
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
* Fix error in diff html rendering
Was missing an optional whitespace check in regex. Also noticed a rare case where diff.Type == Equal would be empty and thus get a newline attached. Fixed that too.
Fixes #13177
* Update services/gitdiff/gitdiff.go
Co-authored-by: zeripath <art27@cantab.net>
* Update gitdiff_test.go
* fmt
Co-authored-by: zeripath <art27@cantab.net>
* Fix diff skipping lines
ParsePatch previously just skipped all lines that start with "+++ " or "--- "
and makes no attempt to see these lines in context.
This PR rewrites ParsePatch to pay attention to context and position
within a patch, ensuring that --- and +++ are only skipped if
appropriate.
This PR also fixes several issues with incomplete files.
Fix https://codeberg.org/Codeberg/Community/issues/308
Fix #13153
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Add testcase
Signed-off-by: Andrew Thornton <art27@cantab.net>
* fix comment
* simplify error handling
Signed-off-by: Andrew Thornton <art27@cantab.net>
* never return io.EOF
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lauris BH <lauris@nix.lv>
* Finally fix diff names
#12771 attempted to fix diff by avoiding the git diff line as
it is possible to have an ambiguous line here.
#12254 attempted to fix diff by assuming that names would quoted
if they needed to be and if one was quoted then both would be.
Both of these were wrong.
I have now discovered `--src-prefix` and `--dst-prefix` which
means that we can set this in such a way to force the git diff
to always be unambiguous.
Therefore this PR rollsback most of the changes in #12771 and
uses these options to fix this.
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Update services/gitdiff/gitdiff.go
* Update services/gitdiff/gitdiff.go
* Update modules/repofiles/temp_repo.go
* fix test
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lauris BH <lauris@nix.lv>
* [Enhancement] Allow admin to merge pr with protected file changes
As tilte, show protected message in diff page and merge box.
Signed-off-by: a1012112796 <1012112796@qq.com>
* remove unused ver
* Update options/locale/locale_en-US.ini
Co-authored-by: Cirno the Strongest <1447794+CirnoT@users.noreply.github.com>
* Add TrN
* Apply suggestions from code review
* fix lint
* Update options/locale/locale_en-US.ini
Co-authored-by: zeripath <art27@cantab.net>
* Apply suggestions from code review
* move pr proteced files check to TestPatch
* Call TestPatch when protected branches settings changed
* Apply review suggestion @CirnoT
* move to service @lunny
* slightly restructure routers/private/hook.go
Adds a lot of comments and simplifies the logic
Signed-off-by: Andrew Thornton <art27@cantab.net>
* placate lint
Signed-off-by: Andrew Thornton <art27@cantab.net>
* skip duplicate protected files check
* fix check logic
* slight refactor of TestPatch
Signed-off-by: Andrew Thornton <art27@cantab.net>
* When checking for protected files changes in TestPatch use the temporary repository
Signed-off-by: Andrew Thornton <art27@cantab.net>
* fix introduced issue with hook
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Remove the check on PR index being greater than 0 as it unnecessary
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: techknowlogick <matti@mdranta.net>
Co-authored-by: Cirno the Strongest <1447794+CirnoT@users.noreply.github.com>
Co-authored-by: zeripath <art27@cantab.net>
Co-authored-by: techknowlogick <techknowlogick@gitea.io>
Following further testing it has become apparent that the diff line
cannot be used to determine filenames for diffs with any sort of predictability
the answer therefore is to use the other lines that are provided with a diff
Fix #12768
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Fix diff path unquoting
services/gitdiff/gitdiff.go whereby there it assumed that the path would
always be quoted on both sides
This PR simplifies the code here and uses fmt.Fscanf to parse the
strings as necessary.
Fix #12546
Signed-off-by: Andrew Thornton <art27@cantab.net>
* Add testcase as per @mrsdizzie
Signed-off-by: Andrew Thornton <art27@cantab.net>
Make previous fix from #12238 more robust since I saw a case where a diff changes only a single character in a chroma class instead of the entire thing. Add another more complicated test to match.
Co-authored-by: Lauris BH <lauris@nix.lv>
* Fix Syntax highlight for token change in added/deleted code
For diffs we first syntax highlight the before and after line then use a 3rd party diff library to find the difference in them and create a substring based on that, which we then highlight a 2nd time to show the specific difference within a line that has changed. In a specific case if the diffrence also changes the chroma class it will split in the middle of the attr and cause broken HTML:
```
<span class="nx">oldtext<span>
<span class="k">var newtext<span>
```
Will then split on
```
<span class="
```
Where the difference starts, and produce something broken like:
```
<span class="<span class="removed-code">nx"oldtext</span></span
```
Fix that by detecting when it happens and putting the HTML back together properly before highlighting the added/deleted code.
Fixes #12235
* fix lint
* apply fix to all diff sections. Also handle case where insert/remove starts with a closing span
* Add a test for this new code
* remove comment
Co-authored-by: Lauris BH <lauris@nix.lv>
Make sure to end up with the same syntax highlighting inside various sections of diffs by processing the code first before detecting specific changes between the lines. Also try and make sure that when highlighting individual lines in a diff that it is tokenized the same as it would be when part of an entire file with more context.
Fixes: #12190
* Make copy/paste work for source code
Fix regression casued by #12047 so copy/paste works properly in all browsers.
Fixes #12184
Also while looking at this I saw a small display issue for blame view. I think #12023 was merged into original PR through an update branch before #12047 was merged and made one of the css ruules not apply anymore.
* use pseudo-element to prevent copying of comment + symbol even when not visually selected
* remove added newline here should not be necessary anymore
* make sure empty line is newline so there is something to select and copy
* Server-side syntax hilighting for all code
This PR does a few things:
* Remove all traces of highlight.js
* Use chroma library to provide fast syntax hilighting directly on the server
* Provide syntax hilighting for diffs
* Re-style both unified and split diffs views
* Add custom syntax hilighting styling for both regular and arc-green
Fixes #7729
Fixes #10157
Fixes #11825
Fixes #7728
Fixes #3872
Fixes #3682
And perhaps gets closer to #9553
* fix line marker
* fix repo search
* Fix single line select
* properly load settings
* npm uninstall highlight.js
* review suggestion
* code review
* forgot to call function
* fix test
* Apply suggestions from code review
suggestions from @silverwind thanks
Co-authored-by: silverwind <me@silverwind.io>
* code review
* copy/paste error
* Use const for highlight size limit
* Update web_src/less/_repository.less
Co-authored-by: Lauris BH <lauris@nix.lv>
* update size limit to 1MB and other styling tweaks
* fix highlighting for certain diff sections
* fix test
* add worker back as suggested
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Lauris BH <lauris@nix.lv>
Unfortunately #11614 introduced a bug whereby the initial commit of a
repository could not be seen due to there being no parent commit to
create a clear diff from.
Here we create a diffstat from the difference between the parentless SHA and the SHA of the empty tree - a constant known to git. (With thanks to @L0veSunshine for informing me of this SHA)
Thanks to @a1012112796 for initial attempt to fix.
Fix #11650
Closes #11674
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-Authored-By: L0veSunshine <xuan199651@gmail.com>
* Some refactor on git diff and ignore getting commit information failed on migrating pull request review comments
* fix test
* fix lint
* Change error log to warn
* fix typo
* Migrate reviews when migrating repository from github
* fix lint
* Added test and migration when external user login
* fix test
* fix commented state
* Some improvements
* fix bug when get pull request and ref original author on code comments
* Fix migrated line; Added comment for review
* Don't load all pull requests attributes
* Fix typo
* wrong change copy head
* fix tests
* fix reactions
* Fix test
* fix fmt
* fix review comment reactions
* Graceful: Create callbacks to with contexts
* Graceful: Say when Gitea is completely finished
* Graceful: Git and Process within HammerTime
Force all git commands to terminate at HammerTime
Force all process commands to terminate at HammerTime
Move almost all git processes to run as git Commands
* Graceful: Always Hammer after Shutdown
* ProcessManager: Add cancel functionality
* Fix tests
* Make sure that process.Manager.Kill() cancels
* Make threadsafe access to Processes and remove own unused Kill
* Remove cmd from the process manager as it is no longer used
* the default context is the correct context
* get rid of double till
In investigating #7947 it has become clear that the storage component of go-git repositories needs closing.
This PR adds this Close function and adds the Close functions as necessary.
In TransferOwnership the ctx.Repo.GitRepo is closed if it is open to help prevent the risk of multiple open files.
Fixes #7947
