Improve repository finder search URL validation

components/dashboard/src/data/git-providers/unified-repositories-search-query.test.ts

@@ -5,7 +5,7 @@
  */
 
 import { SuggestedRepository } from "@gitpod/public-api/lib/gitpod/v1/scm_pb";
-import { deduplicateAndFilterRepositories } from "./unified-repositories-search-query";
+import { deduplicateAndFilterRepositories, isValidGitUrl } from "./unified-repositories-search-query";
 
 function repo(name: string, project?: string): SuggestedRepository {
     return new SuggestedRepository({
@@ -95,3 +95,27 @@ test("it should return all repositories without duplicates when excludeProjects
     expect(deduplicated[0].repoName).toEqual("foo");
     expect(deduplicated[1].repoName).toEqual("bar");
 });
+
+test("should perform weak validation for git URLs", () => {
+    expect(isValidGitUrl("a:")).toEqual(false);
+    expect(isValidGitUrl("a:b")).toEqual(false);
+    expect(isValidGitUrl("https://b")).toEqual(false);
+    expect(isValidGitUrl("https://b/repo.git")).toEqual(false);
+    expect(isValidGitUrl("https://b.com/repo.git")).toEqual(true);
+    expect(isValidGitUrl("[email protected]:")).toEqual(false);
+    expect(isValidGitUrl("[email protected]:")).toEqual(false);
+    expect(isValidGitUrl("[email protected]:22:")).toEqual(false);
+    expect(isValidGitUrl("[email protected]:g/g")).toEqual(true);
+
+    // some "from the wild" cases
+    expect(isValidGitUrl("https://github.com/gitpod-io/gitpod/pull/20281")).toEqual(true);
+    expect(isValidGitUrl("https://gitlab.com/filiptronicek/gitpod.git")).toEqual(true);
+    expect(isValidGitUrl("[email protected]:gitpod-io/gitpod.git")).toEqual(true);
+    expect(isValidGitUrl("[email protected]:filiptronicek/gitpod.git")).toEqual(true);
+    expect(isValidGitUrl("ssh://[email protected]:12345/~/repository.git")).toBe(true);
+    expect(isValidGitUrl("https://bitbucket.gitpod-dev.com/scm/~geropl/test-user-repo.git")).toBe(true);
+    expect(isValidGitUrl("git://gitlab.com/gitpod/spring-petclinic")).toBe(true);
+    expect(isValidGitUrl("[email protected]:v3/services-azure/open-to-edit-project2/open-to-edit-project2")).toBe(
+        true,
+    );
+});

components/dashboard/src/data/git-providers/unified-repositories-search-query.ts

@@ -11,6 +11,7 @@ import { useMemo } from "react";
 import { useListConfigurations } from "../configurations/configuration-queries";
 import type { UseInfiniteQueryResult } from "@tanstack/react-query";
 import { Configuration } from "@gitpod/public-api/lib/gitpod/v1/configuration_pb";
+import { parseUrl } from "../../utils";
 
 export const flattenPagedConfigurations = (
     data: UseInfiniteQueryResult<{ configurations: Configuration[] }>["data"],
@@ -125,17 +126,62 @@ export function deduplicateAndFilterRepositories(
     }
 
     if (results.length === 0) {
-        try {
-            // If the searchString is a URL, and it's not present in the proposed results, "artificially" add it here.
-            new URL(searchString);
+        // If the searchString is a URL, and it's not present in the proposed results, "artificially" add it here.
+        if (isValidGitUrl(searchString)) {
+            console.log("It's valid man");
             results.push(
                 new SuggestedRepository({
                     url: searchString,
                 }),
             );
-        } catch {}
+        }
+
+        console.log("Valid after man");
     }
 
     // Limit what we show to 200 results
     return results.slice(0, 200);
 }
+
+const ALLOWED_GIT_PROTOCOLS = ["ssh:", "git:", "http:", "https:"];
+/**
+ * An opionated git URL validator
+ *
+ * Assumptions:
+ * - Git hosts are not themselves TLDs (like .com) or reserved names like `localhost`
+ * - Git clone URLs can operate over ssh://, git:// and http(s)://
+ * - Git clone URLs (both SSH and HTTP ones) must have a nonempty path
+ */
+export const isValidGitUrl = (input: string): boolean => {
+    const url = parseUrl(input);
+    if (!url) {
+        // SSH URLs with no protocol, such as [email protected]:gitpod-io/gitpod.git
+        const sshMatch = input.match(/^\w+@([^:]+):(.+)$/);
+        if (!sshMatch) return false;
+
+        const [, host, path] = sshMatch;
+
+        // Check if the path is not empty
+        if (!path || path.trim().length === 0) return false;
+
+        if (path.includes(":")) return false;
+
+        return isHostValid(host);
+    }
+
+    if (!url) return false;
+
+    if (!ALLOWED_GIT_PROTOCOLS.includes(url.protocol)) return false;
+    if (url.pathname.length <= 1) return false; // make sure we have some path
+
+    return isHostValid(url.host);
+};
+
+const isHostValid = (input?: string): boolean => {
+    if (!input) return false;
+
+    const hostSegments = input.split(".");
+    if (hostSegments.length < 2 || hostSegments.some((chunk) => chunk === "")) return false; // check that there are no consecutive periods as well as no leading or trailing ones
+
+    return true;
+};

components/dashboard/src/utils.ts

@@ -214,7 +214,7 @@ export class ReplayableEventEmitter<EventTypes extends EventMap> extends EventEm
     }
 }
 
-function parseUrl(url: string): URL | null {
+export function parseUrl(url: string): URL | null {
     try {
         return new URL(url);
     } catch (_) {