The Anonymize form's preview built the readme baseUrl as
"https://github.com/<owner>/<repo>/raw/<source.branch>/". When the
form rendered before the branch field had populated (initial load,
or while waiting on getBranches), the URL became ".../raw//" and
the browser collapsed the empty segment, fetching ".../raw/<file>"
instead of ".../raw/<branch>/<file>". Relative <img src="./X">
references then 404'd against a path with no branch — exactly the
"branch missing" pattern in #407.
Fall back to details.defaultBranch (then "main") so the base URL is
always well-formed.
Fixes#407.
The Anonymize form used the cached RepositoryModel for hasPage,
defaultBranch, etc. — so enabling GitHub Pages (or changing the
default branch) on the source after first cache wouldn't reflect in
the UI, leaving the GitHub Pages checkbox grayed out.
Pass force=1 when loading the form's repo details so the backend
re-queries the GitHub API once. The cost is a single GET /repos/...
call per form load.
Fixes#364.
When the anonymizer doesn't change a slice's text, the streamer used
to push Buffer.from(out, "utf8") — which loses any invalid-UTF-8 bytes
in the input (replaced by U+FFFD via StringDecoder). Files
mistakenly classified as text (binary blobs without a known extension,
text with stray non-UTF-8 bytes, BOMs) came out corrupted even though
nothing in the term list matched.
Track the raw chunk bytes alongside the decoded `pending`. On flush —
where we have every byte buffered — emit the original buffer directly
when the output equals the input, so a pure passthrough is bit-exact.
In the streaming OVERLAP path, do the same when the decode for that
slice round-trips losslessly; fall back to encoded output otherwise
(unchanged from before for that case).
Also add the "missing_content" locale entry for the
/api/anonymize-preview route.
The .leftCol-foot was hidden by default and only revealed inside the
mobile/tablet media query, so the last-update line never appeared on
wider viewports.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two regressions stacked from the recent tree work:
1. expandAllFolders (#496) was marking every folder open, including
folders whose children weren't fetched yet. The directive then
rendered an empty <ul> after each <a>, and the openFolder handler's
"no sibling means we need to load" check silently treated the empty
<ul> as already-loaded — so clicking the folder toggled the class
but the children never appeared.
Skip folders with empty children when pre-expanding, and harden the
click handler so an empty <ul> still triggers a fetch.
2. The $routeUpdate handler (#510 follow-up) became async and called
$scope.$apply(updateContent) at the end. Inside an already-running
digest cycle this no-ops or throws, leaving file navigation stuck.
Run updateContent() synchronously like before, and kick off any
missing parent-directory fetches in the background — getContent()
already falls back to sha "0" when the metadata isn't loaded yet.
Clicking a markdown link into a subdirectory's README threw
"Cannot read properties of undefined (reading 'sha')" and left the
viewer on Loading…. The route handler called updateContent() without
loading the new directory's file listing, so getSelectedFile() returned
undefined and getContent() then dereferenced fileInfo.sha.
Two fixes:
- getContent() falls back to sha "0" when fileInfo is undefined.
- The $routeUpdate handler walks the new path and loads any directory
listings that aren't yet in $scope.files before rendering, so the
selected file actually has its sha by the time we fetch.
Fixes#510.
The form's live README/PR preview was running its own copy of
ContentAnonimizer in the browser. The two implementations had been
drifting — recent fixes for word boundaries (#175/#249), accent
matching (#280), custom replacements (#285), and the diacritic-stripped
variants only landed on the server. Reviewers saw one anonymization;
authors composing the form saw another.
Add POST /api/anonymize-preview that takes a snippet (or a batch) plus
the user's options and runs them through the same ContentAnonimizer
the file route uses. Replace the client-side anonymizeReadme() body
with a debounced call to that endpoint. The PR view's
anonymizePrContent() runs as a synchronous template expression, so it
now reads from a {original -> anonymized} cache that's refreshed in
the background whenever the PR details, terms, or options change.
Single-flight + debounce keep the form responsive; an in-flight
request is dropped on the next change.
Entering an IP address (e.g. 192.168.1.1) or any term with regex
metacharacters made the form invalid because the "regex characters
detected" hint was wired up via $setValidity('terms', 'regex', false).
The text in the UI labels it as a warning, but the form treated it as
an error and refused to save.
Track the warning as a plain $scope flag and show it via ng-show on
that flag, so the form stays valid (#430).
A term entered as "Anonymous=>ABC" now scrubs "Anonymous" to "ABC"
instead of "XXXX-N". Lets users keep anonymized identifiers valid in
source code (no hyphen) and align tokens between paper text and repo.
Indexing for default-mask terms is unchanged: a list of
"Alpha=>AAA", "Beta" still produces XXXX-2 for Beta.
Fixes#285.
The file tree opened collapsed, requiring the reviewer to click each
folder before they could see what was inside. Walk the tree on first
render and mark every folder open in $scope.opens. Folders the user
has explicitly toggled (a previous entry already exists in
$scope.opens) are left as-is, so collapsing still works.
Fixes#496.
Replace the outdated user-images.githubusercontent.com screenshot in
the README with a locally hosted image, and regenerate the in-app
screenshots to reflect the current UI.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The viewer already supported jumping to a line via #L42 in the URL but
never produced one — users had to type it manually. Wire guttermousedown
on the ACE editor to replaceState a #L<n> hash, with shift-click for a
range. Also reapply the highlight on hashchange so pasting a URL into
the address bar works without reload.
Fixes#392.
urlRel2abs() prepended an extra "." when it saw "./X", turning the
relative path into "../X" and silently moving up a directory. As a
result, raw HTML <img src="./imgs/run.png"> inside a README rendered
under /r/<repo>/<file> resolved to /r/<repo>/imgs/... instead of
/r/<repo>/<dir>/imgs/..., so the image 404'd. Markdown image syntax
went through marked-base-url and was unaffected.
Strip the leading "./" instead so the relative path concatenates
cleanly with baseUrl.
Fixes#346.
marked v12 dropped the headerIds option, so headings rendered with no
id attributes and links like [Releases](#releases-and-contributing)
silently failed to scroll. Add a heading renderer that emits a
GitHub-style slug id, with a numeric suffix for duplicates within a
document.
Fixes#390.
The Generate button silently no-op'd because ng-model="newTokenName" inside
an ng-if block wrote to a child scope, leaving $scope.newTokenName empty.
Use $scope.tokenForm.{name,plaintext} so prototypal lookup resolves to the
controller scope.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Toasts used class="toast show" with ng-repeat but never initialized
Bootstrap's toast plugin, so they stayed pinned until manually closed.
Each navigation re-fired toasts (e.g. "README not found in github
pages"), stacking duplicates.
Add an $scope.addToast helper that schedules removal via $timeout, and
route all push sites through it.
Fixes#246.
The Remove action in the dashboard dropdown was gated on
status == 'ready', so expired repos showed no way to be removed and
stuck on the front page. The backend DELETE route already accepts any
non-'removed' status, so widen the ng-show to include 'expired' and
'error'.
Fixes#463.
* Fix horizontal overflow causing page content to be cut off on mobile
- Add overflow-x: hidden to html/body and ng-view to prevent horizontal
scrolling across all pages
- Restore .container.page mobile padding to 15px to match Bootstrap .row
negative margins (-15px), which previously caused 5px overflow per side
- Add max-width: 100% constraints to prevent content from exceeding viewport
https://claude.ai/code/session_01L2xhJCKkjghMDBuwXpSHzi
* Fix Ko-fi widget overlapping hamburger menu and simplify desktop layout
- Move Ko-fi "Support me" button from top-right (where it hid the navbar
hamburger) to bottom-right corner
- Remove card styling (border, background, border-radius) from dashboard
quota section for a flatter, cleaner look
- Remove fixed max-width: 960px from dashboard-page so it uses Bootstrap's
standard container widths, consistent with other pages like admin
https://claude.ai/code/session_01L2xhJCKkjghMDBuwXpSHzi
* Redesign anonymize page: centered landing, reorganized form layout
- No URL state: centered input in the middle of the page for a clean
initial experience
- URL provided state: preview on the left, form settings on the right
in a fixed-width sidebar panel (380px on desktop)
- Reorganized form sections into logical groups:
Source (branch/commit) → Identity (ID/conference) → Anonymization
(terms) → Display (checkboxes, no longer hidden in accordion) →
Expiration
- Removed card/accordion wrappers for a flatter, more scannable form
- Mobile: form stacks below preview with sticky submit bar
https://claude.ai/code/session_01L2xhJCKkjghMDBuwXpSHzi
* Reduce navbar height on desktop
- Reduce navbar padding from default .5rem to 4px vertical
- Shrink nav icons from 30px/40px to 20px/28px
- Reduce nav-link font size to 0.9rem with tighter padding
- Shrink navbar-brand font size to 1rem
https://claude.ai/code/session_01L2xhJCKkjghMDBuwXpSHzi
---------
Co-authored-by: Claude <noreply@anthropic.com>
Compatibility with GitHub Pages is limited: Jekyll (and other static
site generators) are not supported. This change documents this
limitation on the home page and FAQs.
Although Markdown files are converted to HTML and thus accessible when
anonymized GitHub Pages is enabled for the repository, the Markdown to
HTML conversion includes only the content -- i.e., there is no special
formatting (as would be available when using Jekyll).
For anyone who wants to use Jekyll, a potential workaround is to build
the site locally and commit the generated site to the repository.
Closes#269
Co-authored-by: Joel Coffman <joel.coffman@acm.org>