Repository navigation
chore(wiki): correct stale SolidPing facts in the competitor comparison - #319
Merged
Merged
Conversation
Writing the public /compare pages against the code surfaced four claims in this wiki that no longer match the implementation, plus one missing rule. - Browser checks are **chromedp over CDP**, not Rod. Rod was the engine the 2026-01-05 screenshot idea proposed; the shipped code (server/internal/checkers/checkbrowser) uses chromedp. Corrected in maintenant.md, betterstack/comparison.md, comparison/monitor-types.md and comparison/solidping-position.md — the website had inherited "Rod / headless Chromium" from here. - Screenshot capture: browser checks do capture on failure since spec 2026-08-21-01. Only HTTP-failure capture is still missing. - The alerting table listed Microsoft Teams, Telegram and PagerDuty as gaps with "spec ready". All three ship. Matrix too. Total channels was "10 native"; the registry has 16 senders. - SMS/voice was scored a flat gap. It is implemented and works self-hosted on the operator's own Twilio account; what is true is that it is **not included on any hosted plan**, because carriers bill per message. That distinction decides whether a comparison should score it as a loss, so it is now spelled out rather than flattened to an X. Adds a "Which SolidPing is being compared" section: self-hosted against self-hosted tools, SolidPing Cloud with its real caps against hosted services, both for tools that sell both. The existing "Advantages Over All Three" table wins every row precisely because it compares our free build to their paid plans, and now says so. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
fclairamb
added a commit
that referenced
this pull request
Sep 4, 2026
## Why `wiki/` holds internal engineering notes and is an input to **no build**. The published docs site is `web/docs/`, which is separate; every other reference to `wiki/` in the tree is a comment pointing at it. So a wiki-only pull request currently spends a ~28-minute E2E run, a Go build, four frontend builds and a full backend test suite proving that prose did not break the compiler. That is not hypothetical — #319 (five files, all under `wiki/`) did exactly that. ## What changed A `changes` job asks the API which files the PR touches and sets one `code` output. The seven build jobs are gated on it. **Nothing else moves** — the jobs, their steps and their ordering are untouched. This is safe *specifically* because of how the existing `ci` gate is written: `always()`, failing only on `failure` or `cancelled`, so a skipped job already satisfies it. That is the same property the release-please skip has relied on all along, and `main` has no required status checks, so there is no check that can hang unreported. If that gate ever starts treating `skipped` as a failure, this must change with it — there is a comment saying so. The gate now also depends on `changes` itself. Without that, a detector that *failed* would skip every build job and leave the gate looking at an all-skipped graph — turning a broken detector into a way to merge untested code. ## Fail-safes The failure mode here is silently not testing things, so the decision fails toward building: - any non-`pull_request` event (pushes to `main`, in particular) always builds; - an empty file list from the API means "the API told us nothing", not "nothing changed" — it always builds; - the wiki-only decision is a POSIX `case` glob, not `grep -qv '^wiki/'`. That last point is not theoretical. `ugrep`, which is `grep` on some developer machines, returns false for inverted-quiet matching even when non-wiki files are present: it classified a **138-file code change as "wiki-only"**, while `grep -cv` on the same input correctly counted 126 non-wiki files. GNU grep on the runner would have behaved correctly, but a check whose meaning depends on which `grep` is installed is not one worth shipping. ## Verification Both live PRs, evaluated with the exact decision logic from the workflow: | PR | Files | Result | |---|---|---| | #319 | 5, all under `wiki/` | `code=false` → skips | | #318 | 138 | `code=true` → builds | YAML parses, and the resulting job graph was checked: all seven build jobs gain `changes` in `needs`, and the `ci` gate depends on it too. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Found while rewriting the public
/comparepages (solidping-website#28) against the code rather than against this wiki. Four claims here no longer match the implementation, and the website had inherited one of them verbatim.server/internal/checkers/checkbrowser,server/go.mod2026-08-21-01,checkbrowser/config.goserver/internal/notifications/registry.go,db/models/user_contact.goentitlements/defaults.go(MaxSmsPerMonth0) + the website's/pricingcopyThat last one is the interesting one: flattening it to an X loses the distinction that decides how a comparison should score it. Against Better Stack, UptimeRobot, Pingdom, StatusCake or Site24x7 on hosted terms it is a real loss. Against Uptime Kuma or Gatus on self-hosted terms it is not a loss at all. Both are now stated.
Also adds a "Which SolidPing is being compared" section at the top of
comparison/solidping-position.md, with the plan ladder and the self-hosted entitlements side by side and pointers to the two files they come from. The existing "Advantages Over All Three" table wins every single row because it puts our free uncapped build against their paid plans — that is now labelled rather than left to look like a finding.Docs only.
🤖 Generated with Claude Code