// Package chimw is the chi half of the shared HTTP middleware for the custom // services of a self-hosted SourceHut instance (compare, spec, dolt, cover, // bench, tokens). // // It is a second middleware package because the first one may not grow these. // sr-ht-ecore/middleware imports nothing but net/http, on purpose: it is the // package a handler, an API surface or a plain http.ServeMux can reach for // without inheriting anybody's router. That rule is what kept the donors' getHead // out of it — three lines, all three of them chi's — and the rule is right. It is // not, however, an argument for leaving the helper copied into every router that // wants it. Every custom service on this instance routes with chi; a package // that says so in its name costs a service that already imports chi exactly // nothing, and costs middleware's rule nothing either. // // The boundary between the two is one question: does the helper have to know // what a route is? A cache header, a panic guard and a status constant do not, // and live in middleware. Registering a pair of methods for one pattern, // installing the two handlers chi calls when routing fails, and reading the // request id chi's RequestID middleware put in the context all do, and live // here. // // Three things, each of them either copied identically by the services that have // it or absent from the ones that do not — which is the same finding twice over: // // - GetHead, the read route registered under GET and HEAD both. Five copies: // cover, bench and tokens on their web surface, cover and bench again on // their API, where the same three lines carry a different name each time. // - RenderRefusals, chi's NotFound and MethodNotAllowed pointed at the // service's own error page. spec and dolt install neither, so an unrouted GET // there answers with net/http's plain-text 404 while every other refusal on // the instance is a rendered page. // - RequestLogger, the request line as a slog record instead of chi's // stdlib-log line on stdout. // // Usage, in the order a service installs them: // // r.Use(chimiddleware.RequestID) // r.Use(chimiddleware.RealIP) // r.Use(chimw.RequestLogger(chimw.SlogFormatter{Skip: chimw.SkipPaths("/healthz")})) // r.Use(middleware.RecoverPanics(func(w http.ResponseWriter, r *http.Request, _ any) { // s.renderError(w, r, http.StatusInternalServerError, pages.InternalMessage) // })) // r.Use(middleware.PrivateCache) // // chimw.RenderRefusals(r, s.renderError) // chimw.GetHead(r, "/healthz", s.handleHealthz) // // The daemons that adopt this will have to rename an import: every one of them // already aliases chi's own middleware package as chimw, which is exactly the // name this one takes by default. chimiddleware, as spelled above and in this // package's own files, is the alias to move them to — the shorter name belongs // to the package a service writes against. // // The request logger is the one middleware that belongs outside // middleware.RecoverPanics rather than inside it. It has to observe the status // that actually went out, and for a panicking handler that status is the 500 // RecoverPanics renders — a logger installed underneath would see the unwinding // stack and nothing written, and would log the request that produced the error // page as if it had produced nothing. chi's own Logger carries the same // instruction for the same reason. Nothing is left uncovered by moving it out: // this middleware builds a record and calls slog, and a panic in slog is not a // failure any error page is going to survive either. // // Records go to slog's default logger unless one is named, exactly as // middleware.RecoverPanics does and for the same reason: a library has no // business choosing a handler. The service installs its own — scribe's tinted // one on this instance — with slog.SetDefault at startup, and the request lines // land in the same stream, with the same masking rules, as everything else it // logs. package chimw import ( "net/http" "github.com/go-chi/chi/v5" "sourcecraft.dev/bigbes/sr-ht-ecore/pages" ) // GetHead registers one read route under GET and HEAD both. // // chi's Get registers GET alone, and net/http synthesizes nothing for a custom // handler, so every read route of a donor answered `curl -I` — and every uptime // probe, and every cache revalidating what it holds — with a 405 naming GET as // the only method it would accept, plus a kilobyte of rendered error page from a // path whose whole job is to say "yes" cheaply. RFC 9110 §9.3.2 makes HEAD // mandatory for any resource that serves GET, and HEAD is the one method a // monitor reaches for precisely because it costs no body. // // The two share the handler rather than getting one each, and that is the whole // design: nothing in these services looks at r.Method on a read path, so a HEAD // produces exactly the status line and the headers its GET would have, computed // by the same code on the same query. There is no second path that could answer // 200 where the GET answers 404, which on these pages would be a visibility leak // (SPEC ch. 6.3). The body is dropped by net/http, which discards writes on a // HEAD response after counting them, so the Content-Length a client gets is the // real one. // // It is this and not chi's own middleware.GetHead, which registers nothing and // instead rewrites, per request, the method the routing tree is walked with. // // The donors' comments say that rewrite lands on r.Method, so that every line // written about the request — the log record of RequestLogger below, the panic // report of middleware.RecoverPanics — names a method the viewer did not send. // Read chi v5's source and it does not: it sets RouteMethod on the route context // and leaves the request alone. The correction is worth keeping written down, // because what is left is the half of the objection that no ordering or logging // discipline can work around. // // The tree stops describing the service. A route registered for GET alone is a // route that does not serve HEAD as far as anything reading the tree is // concerned — a chi.Walk, which is how the donors' TestEveryGetRouteHasAHeadTwin // asks the question at all, and chi's own 405 handler, which builds its Allow // header out of the methods that were registered. The service answers HEAD // anyway, from a middleware, and the two records disagree; the request context // then says GET while the request says HEAD, so which one a handler or a later // middleware believes depends on which it happened to read. Registering the pair // leaves one record of what is served, and it is the tree. // // It is also two lines against a middleware that re-enters routing on every HEAD // that missed. // // Mutating routes deliberately do not go through it. A HEAD that writes is not a // HEAD, so a form's POST is registered with r.Post next to its GET, and a HEAD // on that path resolves to the page. func GetHead(r chi.Router, pattern string, h http.HandlerFunc) { r.Get(pattern, h) r.Head(pattern, h) } // An ErrorRenderer is a service's own error page, as every donor already spells // it: the status it is answering with and the sentence the viewer can act on. // // It is this shape and not middleware.RecoverPanics's (w, r, recovered) because // the two callbacks are answering different questions. A panic has one status // and one message and hands over a value to classify; a refusal has no value and // two statuses, and the renderer is the donors' existing renderError method, // which is passed by name rather than wrapped in a closure per call site. type ErrorRenderer func(w http.ResponseWriter, r *http.Request, status int, message string) // RenderRefusals points chi's two routing failures at the service's error page. // // A path the router does not serve and a method it does not allow are answered // by chi with net/http's Error: text/plain, no chrome, no nav, no way out for a // viewer who mistyped a URL. Five of the donors' routers install a pair of // closures to fix that and spell them identically, compare installs the 404 // alone and lives with chi's 405, and spec and dolt install neither — so a wrong // address on those two is the only refusal on the instance that does not look // like the service it came from. // // The messages are the shared ones — pages.NotFoundMessage and // pages.MethodMessage — rather than the caller's. The 404 above all is not free // prose: the visibility rules of these services require "somebody else's private // thing" and "no such thing" to be indistinguishable (SPEC ch. 6.3), and a // router's 404 that differed in its wording from a handler's would rebuild by // hand the distinction the status code was chosen to erase. A service that wants // other prose has the seam anyway, in the renderer it passes. // // One thing is lost by taking the 405 over from chi and is worth knowing rather // than discovering: chi hands the list of methods that *would* have matched only // to its own default handler, through unexported types, so a custom one cannot // emit the Allow header RFC 9110 §15.5.6 asks a 405 for. Every donor already // made that trade, silently. It is the right way round for these services, whose // 405s are answered to browsers rather than to clients negotiating a method, and // it is recorded here so that a service that does need Allow knows it has to // walk the tree for it. // // It is a function taking a router, not a middleware, because chi's NotFound and // MethodNotAllowed are registrations on the routing tree and not links in a // chain: they run after routing has failed, so a chain has nothing left to hand // them. That also means they are inherited by every sub-router mounted later — // installing them once on the root is enough. // // A nil renderer is a wiring mistake and panics here, at construction, rather // than at the first mistyped URL. func RenderRefusals(r chi.Router, render ErrorRenderer) { if render == nil { panic("chimw: RenderRefusals needs a render callback") } r.NotFound(func(w http.ResponseWriter, r *http.Request) { render(w, r, http.StatusNotFound, pages.NotFoundMessage) }) r.MethodNotAllowed(func(w http.ResponseWriter, r *http.Request) { render(w, r, http.StatusMethodNotAllowed, pages.MethodMessage) }) }