master
7
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 hermes
|
be8c959de2 |
fix: dialogs rendered EMPTY - relative project path caused a silent 403
THE ACTUAL CAUSE of "codegit is just a black box". Sections 4 and 5 were real bugs but
neither was it; this is.
CHAIN
1. install/process.php stores a RELATIVE folder name -> projects.db {"Blog":"blog"}
2. common.php::loadState() does $projectPath = reset($projectList) -> "blog", and
SESSION("projectPath","blog") PERSISTS it, so the bad value sticks across logins
3. filetree puts it straight into the DOM: <a id="project-root" data-path="blog">
4. codegit sends repo="blog"; getWorkspacePath("blog") fails checkPath() ->
traits/path.php:67 Common::send(403, "Client does not have access.")
5. modal.js SILENTLY discards it:
settled: function(reply,status){ if (status !== 200) return; ... }
No toast, no console error, #content left empty.
That last step is why it presents as an empty black dialog rather than an error - the request
fails and nothing reports it. Which is exactly why it survived the CSS fixes.
PINNED DOWN BY CAPTURING THE REAL REQUEST, not by reading CSS:
POST /dialog.php {"target":"codegit","action":"codegit","repo":"blog"} -> 403
#content innerHTML length: 0
getWorkspacePath("blog") -> 403
projects store -> {"Blog":"blog"}
FIX (scripts/fix-project-path.sh) - both parts required:
A. resolve the path in loadState and re-store it, which also HEALS sessions already
holding the relative value (and re-persists so it stays healed)
B. make the projects store absolute, so array_search() can still map path -> name for
the tree label. Fixing only (A) would silently break the label: the path would resolve
but array_search would return false and the tree would show an empty project name.
VERIFIED from a deliberately poisoned session ("blog"): it self-heals to the absolute path,
dialog returns 200, #content fills with 1758 chars. An EXISTING browser session heals on its
next page load - no hard refresh or re-login needed.
Tests: 47 passed, 0 failed, 1 accepted-risk
+ projects store must hold an absolute path
+ loadState must resolve a relative projectPath
Also marks section 4 as superseded so a future reader doesn't stop at the CSS explanation.
|
||
|
hermes
|
48e439ed6e |
fix: cache-bust asset URLs - the real reason the panel fix appeared not to work
Reported again as "code git still no go just a black box is it because I'm on firefox?"
It is NOT Firefox. Tested properly: installed Firefox 155 and drove the live server with
Playwright, authenticated. Result:
assets Firefox requested: theme/main.min.css?v=1790663824 <- versioned
applied panel rule: height: auto; max-height: 58vh <- the fix IS applied
dialog 193px | panel 138px | toolbar INSIDE | JS errors: none
CodeGit renders correctly in Firefox. Firefox was simply still running the OLD CSS.
WHY THE EARLIER FIX DIDN'T TAKE
Atheos emits asset URLs with no version:
<link rel="stylesheet" href="theme/main.min.css">
<script src="public/modules.min.js"></script>
Same path before and after a change, and originally no cache headers at all. So the browser
keeps its cached copy and never even asks the server. Changing the CSS server-side cannot
help when the browser isn't fetching it - which is why the previous Cache-Control fix was
necessary but NOT sufficient against a cache that was already poisoned.
FIX - versioned URLs
SourceManager::getTag() and the theme link in index.php now append the file mtime:
theme/main.min.css?v=1790663824
A changed file gets a NEW URL, so it is fetched immediately; an unchanged file keeps its URL
and stays cached. This works even against an already-stale cache, which headers alone cannot
do. The page itself is served no-store, so the next normal reload picks up new URLs
automatically - no hard refresh needed by the user.
Both parts are now in scripts/fix-panel-layout.sh (idempotent, verified):
1. Cache-Control: no-cache for *.css/*.js (revalidation; keeps 304s fast)
2. mtime-versioned asset URLs (the part that actually defeats a poisoned cache)
Tests: 45 passed, 0 failed, 1 accepted-risk
+ asset URLs must carry ?v= (guards against this regressing and silently hiding future fixes)
|
||
|
hermes
|
c57aff3c33 |
docs: branch protection is a recommendation, not a blocker (operator decision)
The operator has decided not to enable branch protection on armistace/blog while the editor is LAN-only: "leave it as a recommendation - I won't do that unless I take this out to the wider internet" That is a reasonable call, so this records it as an ACCEPTED RISK rather than an open task. Test behaviour changed to match. The suite still attempts a direct push to master with the deploy key on EVERY run - the check is not removed, because that is what would catch the setting being lost or never applied. But when protection is off it now reports ACCEPTED-RISK master is NOT protected - the deploy key CAN push master directly instead of FAIL, so the suite is green and CI-usable (it previously exited 1, meaning a genuine regression would have been buried in expected noise). The run flips back to "PASS master is protected" by itself the moment protection is enabled, so the recommendation self-cancels with nobody having to remember it. Summary line now reads RESULT: 44 passed, 0 failed, 1 accepted-risk Docs record WHY the risk is acceptable (the compensating control is reachability, not immutability - nothing off-LAN can reach the editor, and the only LAN user is the operator) and the two triggers that turn it back into a real finding: 1. the editor becomes reachable off-LAN (NPM proxy, port-forward, VPN route) 2. anyone other than the operator gets access Written down because an accepted risk is only acceptable if it is explicit and has a revisit trigger - otherwise it just looks like an oversight later. |
||
|
hermes
|
2d57173e31 |
fix: CodeGit looked like an empty black box - hardcoded panel height + cache
Reported as "why can't i see anything from codegit its just a black box".
NOT a codegit bug. Verified first, exhaustively, through the authenticated API:
overview 200 940 bytes
log 200 79596 bytes (full commit history)
transfer 200 1460 bytes
config 200 1174 bytes
no JS errors in the console; every DOM helper codegit calls exists.
ROOT CAUSE 1 - a generic CSS rule hardcodes a height
theme/modules/common/_panel.scss -> theme/main.min.css:
panel { ... height: 500px; ... }
That forces EVERY modal panel to a fixed 500px box. Measured live:
#dialog 404px (clamped by max-height:70vh + overflow-y:auto)
#codegit panel 500px, while its content is only 119px
toolbar ("Branch") top 620 vs dialog bottom 490 -> OUTSIDE the modal
So ~380px of the dialog was empty dead space and the Branch line rendered below the
modal, invisible. Fixed to height:auto with max-height:58vh (long content still scrolls):
before: panel 500 -> toolbar top 620, dialog bottom 490 (outside)
after: panel 136 -> toolbar top 256, dialog bottom 279 (inside)
_panel.scss updated too so the source does not contradict the shipped CSS.
ROOT CAUSE 2 - the fix was invisible because of the browser cache
Atheos serves theme/main.min.css and the generated bundles from stable URLs with no cache
headers, so browsers legitimately reuse their cached copy - which is why the first attempt
appeared to do nothing. nginx now sends Cache-Control: no-cache, must-revalidate for
*.css/*.js: caching still works (cheap 304s) but changes are always picked up.
Adds
scripts/fix-panel-layout.sh idempotent re-apply for both fixes (lost on update-atheos)
docs/known-issues.md §4, §5 both findings, with the measurements
Tests (44 passed, 1 failed - the failure is branch protection, a genuine open item)
+ panel must not be fixed at 500px
+ theme CSS must send Cache-Control: no-cache
|
||
|
hermes
|
c4a4de1da7 |
fix: rDelete fataled on musl (GLOB_BRACE) - broke ALL deletion and plugin removal
Reported as "deleting terminal just cycles in the ui". Root cause is much larger than the
plugin: it was a platform bug that broke deletion everywhere.
CAUSE
traits/helpers.php::rDelete() did:
glob($target . "{*,.[!.]*,..?*}", GLOB_BRACE|GLOB_MARK);
GLOB_BRACE is a GLIBC extension. Alpine is musl, where the constant is undefined,
and referencing an undefined constant is a FATAL in PHP 8:
PHP Fatal error: Uncaught Error: Undefined constant "GLOB_BRACE"
in traits/helpers.php:138
#0 components/market/class.market.php(178): Common::rDelete()
#1 components/market/controller.php(46): Market->remove()
The fatal kills the request before Common::send(200), so the front end never gets a
settle response and the UI spins forever. A hang, not an error - which is why it
presented as "cycles in the ui".
BLAST RADIUS (measured, two call sites)
components/market/class.market.php:178 removing a plugin from the market
components/filetree/class.filetree.php:59 DELETING ANY FILE OR FOLDER IN THE EDITOR
The second is core functionality, so this was never a plugin bug - installing the
Terminal plugin is just what made it visible.
FIX
patches/helpers-rDelete-GLOB_BRACE.diff - expand the brace pattern by hand when
GLOB_BRACE is undefined, keeping upstream behaviour where it is defined. Verified:
nested dirs + dotfiles + single files all delete correctly now.
Upstream has NOT fixed this (checked main) so it cannot be resolved by updating.
PLUGIN REMOVED
Once deletion worked, removal via the market succeeded and the plugin is now gone:
{"text":"Plugin successfully removed.","status":200}
That is the right end state - a browser shell was the largest attack surface, and
macros already provide command execution. Removed through the market's own code path
so the plugin list, addons cache and bundle stayed consistent.
TESTS (now 42 passed, 1 failed - the failure is branch protection, a genuine open item)
+ a GLOB_BRACE regression guard that fails loudly if the patch is lost to an update
+ the identity guard is INVERTED: FR-11 required an EMPTY identity so the agent could
not impersonate the operator, but the operator has since asked for it to be set, so
the test asserting emptiness was asserting the opposite of the current design. It
now requires the identity to be present and to NOT look agent-generated.
docs/known-issues.md: new section 1 documents the GLOB_BRACE fault in full, including the
honest note that this is the real cost of choosing Alpine over a glibc base image.
|
||
|
hermes
|
0b72cf48b1 |
test: add a master-protection guard, fix three stale assertions
Responds to the independent review of the writer-branch PR.
Adopted (the reviewer's strongest recommendation): a test that PROVES master is
protected, rather than documenting the setting. It attempts a direct push to
master with the deploy key and requires it to be rejected - the same way an
attacker would find out. It currently FAILS, which is the correct result: branch
protection is genuinely not enabled on armistace/blog, and the failure is now
visible in every test run instead of buried in prose.
Cannot fix from here: setting branch protection on armistace/blog returns
403 "user should be an owner or a collaborator with admin write"
for the bot account (it is a collaborator, not the owner). This is an operator
action, documented in docs/writer-branch-workflow.md with the exact Gitea
settings. Until it is done the writer-branch model is advisory, not enforced.
Also fixed three assertions that were stale rather than wrong:
- "installer markers present" now accepts the post-install state (config.php
populated) as valid, since the operator has completed the installer.
- "posts found == 20" is now ">= 20" - the operator has added a file, so an
exact count was always going to fail.
- "branch == master" is now master|atheos-writer - the writer branch is the
intended steady state under the new workflow, so asserting master was
asserting the opposite of the design.
Result: 40 passed, 1 failed - the single failure being the branch-protection gap.
|
||
|
hermes
|
a6fb61d0ea |
feat: provision Alpine LXC 136 running Atheos, wired to Gitea
Provisions a LAN-only browser code editor (Atheos) in an Alpine 3.23 LXC on the scotty PVE node, with armistace/blog cloned and ready to edit. Built via the AI-DLC workflow; the full process trail is in aidlc-docs/. Deliverables - scripts/provision.sh rebuild the guest from scratch (re-runnable) - scripts/update-atheos.sh move Atheos to latest upstream and RE-PIN - scripts/build-and-test.sh captures evidence for every acceptance criterion - files/atheos.conf nginx vhost (socket path byte-matches the FPM pool) - files/ssh-config ~atheos/.ssh/config deploy-key alias - RUNBOOK.md operational guide (START AT STEP 1) - docs/security-notes.md credential model + wider-web hardening TODOs - docs/resiliency-notes.md what a non-PoC deployment would need Design decisions - D1 no git identity is configured by the agent; the operator sets it. Git will refuse the first commit until then, which is deliberate (requirement FR-11). - D2 Atheos pinned to commit f90cd18 (build 611) with a verified tarball sha256, plus update-atheos.sh as the deliberate escape hatch. - D3 nginx :80, web root /var/www/atheos, php-fpm over a unix socket. - D6 git auth is a per-repository DEPLOY KEY scoped to armistace/blog, not a personal access token (a PAT's write:repository scope reaches every repo the account can write). Verified on the live guest (40 checks, 0 failures) - AC-2 Atheos served over HTTP 200 by nginx+php-fpm; socket paths byte-match - AC-3 20 posts + themes/cleanblog present; owned by the php-fpm user - AC-4 all five writable paths writable as the service user - AC-5 remote targets armistace/blog; deploy key authenticates over SSH - AC-6 survival proven with a real container restart, marker byte-identical - AC-8 no python/pelican/pip/make and no DB engine - scope boundary enforced - FR-11 git identity is empty; deploy key mode 600 owned by the service user Traps documented in the RUNBOOK (each cost real debugging time) - The Gitea SSH user is `gitea`, NOT `git` (app.ini SSH_USER=gitea). With `git`, sshd answers "Invalid user git" and rejects before reading authorized_keys, which is indistinguishable from a bad key. - The Gitea SSH endpoint is the LAN address (:22 is not forwarded publicly). - `uptime` and `free` report the PVE HOST inside an LXC without lxcfs, so a container that just restarted can claim days of uptime. |