feat(chromium): add --chromium-clear-storage to clear local storage between conversions (#919)

.bruno/Chromium/Screenshot/HTML Screenshot.bru

@@ -15,6 +15,7 @@ body:multipart-form {
   ~width: 800
   ~height: 600
   ~clip: false
+  ~selector:
   ~format: png
   ~quality: 100
   ~optimizeForSpeed: false

.bruno/Chromium/Screenshot/Markdown Screenshot.bru

@@ -16,6 +16,7 @@ body:multipart-form {
   ~width: 800
   ~height: 600
   ~clip: false
+  ~selector:
   ~format: png
   ~quality: 100
   ~optimizeForSpeed: false

.bruno/Chromium/Screenshot/URL Screenshot.bru

@@ -15,6 +15,7 @@ body:multipart-form {
   ~width: 800
   ~height: 600
   ~clip: false
+  ~selector:
   ~format: png
   ~quality: 100
   ~optimizeForSpeed: false

.github/workflows/continuous-integration.yml

@@ -31,7 +31,7 @@ jobs:
       - name: Run linters
         uses: golangci/golangci-lint-action@v9
         with:
-          version: v2.10.1
+          version: v2.12.2
 
   lint-prettier:
     name: Lint non-Golang codebase

pkg/modules/chromium/browser.go

@@ -49,6 +49,7 @@ type browserArguments struct {
 	denyPublicIPs     bool
 	clearCache        bool
 	clearCookies      bool
+	clearStorage      bool
 	disableJavaScript bool
 }
 
@@ -222,7 +223,7 @@ func (b *chromiumBrowser) Start(logger *slog.Logger) error {
 		if stopErr != nil {
 			logger.ErrorContext(context.Background(), fmt.Sprintf("stop pinning proxy after failed start: %s", stopErr))
 		}
-		return fmt.Errorf("run exec allocator: %w", err)
+		return fmt.Errorf("run exec allocator: %w; if Chromium is slow to start, raise --chromium-start-timeout (currently %s)", err, b.arguments.wsUrlReadTimeout)
 	}
 
 	b.ctxMu.Lock()
@@ -367,6 +368,7 @@ func (b *chromiumBrowser) pdf(ctx context.Context, logger *slog.Logger, url, out
 		runtime.Enable(),
 		clearCacheActionFunc(logger, b.arguments.clearCache),
 		clearCookiesActionFunc(logger, b.arguments.clearCookies),
+		clearStorageActionFunc(logger, b.arguments.clearStorage, url),
 		disableJavaScriptActionFunc(logger, b.arguments.disableJavaScript),
 		setCookiesActionFunc(logger, options.Cookies),
 		userAgentOverride(logger, options.UserAgent),
@@ -393,6 +395,7 @@ func (b *chromiumBrowser) screenshot(ctx context.Context, logger *slog.Logger, u
 		runtime.Enable(),
 		clearCacheActionFunc(logger, b.arguments.clearCache),
 		clearCookiesActionFunc(logger, b.arguments.clearCookies),
+		clearStorageActionFunc(logger, b.arguments.clearStorage, url),
 		disableJavaScriptActionFunc(logger, b.arguments.disableJavaScript),
 		setCookiesActionFunc(logger, options.Cookies),
 		userAgentOverride(logger, options.UserAgent),

pkg/modules/chromium/chromium.go

@@ -43,6 +43,10 @@ var (
 	// or undefined.
 	ErrInvalidSelectorQuery = errors.New("invalid selector query")
 
+	// ErrScreenshotSelectorNotFound happens when the CSS selector of a
+	// screenshot matches no element with a rendered box.
+	ErrScreenshotSelectorNotFound = errors.New("screenshot selector not found")
+
 	// ErrRpccMessageTooLarge happens when the messages received by
 	// ChromeDevTools are larger than 100 MB.
 	ErrRpccMessageTooLarge = errors.New("rpcc message too large")
@@ -347,6 +351,11 @@ type ScreenshotOptions struct {
 	// dimensions.
 	Clip bool
 
+	// Selector clips the screenshot to the bounding box of the first element
+	// matching this CSS selector. Empty captures the whole page. Takes
+	// precedence over Clip.
+	Selector string
+
 	// Format is the image compression format, either "png" or "jpeg" or
 	// "webp".
 	Format string
@@ -370,6 +379,7 @@ func DefaultScreenshotOptions() ScreenshotOptions {
 		Width:             800,
 		Height:            600,
 		Clip:              false,
+		Selector:          "",
 		Format:            "png",
 		Quality:           100,
 		OptimizeForSpeed:  false,
@@ -467,6 +477,7 @@ func (mod *Chromium) Descriptor() gotenberg.ModuleDescriptor {
 			fs.Bool("chromium-deny-public-ips", false, "Reject URLs whose host resolves to a public IP address. Enable on air-gapped or data-governed deployments to prevent outbound traffic from leaving a private network")
 			fs.Bool("chromium-clear-cache", false, "Clear Chromium cache between each conversion")
 			fs.Bool("chromium-clear-cookies", false, "Clear Chromium cookies between each conversion")
+			fs.Bool("chromium-clear-storage", false, "Clear Chromium local storage between each conversion (session storage is already isolated per conversion)")
 			fs.Bool("chromium-disable-javascript", false, "Disable JavaScript")
 			fs.Bool("chromium-disable-routes", false, "Disable the routes")
 
@@ -518,6 +529,7 @@ func (mod *Chromium) Provision(ctx *gotenberg.Context) error {
 		denyPublicIPs:     flags.MustBool("chromium-deny-public-ips"),
 		clearCache:        flags.MustBool("chromium-clear-cache"),
 		clearCookies:      flags.MustBool("chromium-clear-cookies"),
+		clearStorage:      flags.MustBool("chromium-clear-storage"),
 		disableJavaScript: flags.MustBool("chromium-disable-javascript"),
 	}
 

pkg/modules/chromium/routes.go

@@ -366,6 +366,7 @@ func FormDataChromiumScreenshotOptions(ctx *api.Context) (*api.FormData, Screens
 	var (
 		width, height     int
 		clip              bool
+		selector          string
 		format            string
 		quality           int
 		optimizeForSpeed  bool
@@ -376,6 +377,7 @@ func FormDataChromiumScreenshotOptions(ctx *api.Context) (*api.FormData, Screens
 		Int("width", &width, defaultScreenshotOptions.Width).
 		Int("height", &height, defaultScreenshotOptions.Height).
 		Bool("clip", &clip, defaultScreenshotOptions.Clip).
+		String("selector", &selector, defaultScreenshotOptions.Selector).
 		Custom("format", func(value string) error {
 			if value == "" {
 				format = defaultScreenshotOptions.Format
@@ -420,6 +422,7 @@ func FormDataChromiumScreenshotOptions(ctx *api.Context) (*api.FormData, Screens
 		Width:             width,
 		Height:            height,
 		Clip:              clip,
+		Selector:          selector,
 		Format:            format,
 		Quality:           quality,
 		OptimizeForSpeed:  optimizeForSpeed,
@@ -963,7 +966,17 @@ func screenshotUrl(ctx *api.Context, chromium Api, url string, options Screensho
 	outputPath := ctx.GeneratePath(ext)
 
 	err := chromium.Screenshot(ctx, ctx.Log(), url, outputPath, options)
-	err = handleChromiumError(err, options.Options)
+	if errors.Is(err, ErrScreenshotSelectorNotFound) {
+		err = api.WrapError(
+			err,
+			api.NewSentinelHttpError(
+				http.StatusBadRequest,
+				fmt.Sprintf("The selector '%s' (selector) matched no element with a visible box", options.Selector),
+			),
+		)
+	} else {
+		err = handleChromiumError(err, options.Options)
+	}
 	if err != nil {
 		return fmt.Errorf("screenshot: %w", err)
 	}

pkg/modules/chromium/tasks.go

@@ -6,13 +6,16 @@ import (
 	"errors"
 	"fmt"
 	"log/slog"
+	"net/url"
 	"os"
+	"strconv"
 	"time"
 
 	"github.com/chromedp/cdproto/cdp"
 	"github.com/chromedp/cdproto/emulation"
 	"github.com/chromedp/cdproto/network"
 	"github.com/chromedp/cdproto/page"
+	"github.com/chromedp/cdproto/storage"
 	"github.com/chromedp/chromedp"
 	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/codes"
@@ -54,6 +57,7 @@ func printToPdfActionFunc(reqCtx context.Context, logger *slog.Logger, outputPat
 		defer span.End()
 
 		err := func() error {
+			paperWidth := options.PaperWidth
 			paperHeight := options.PaperHeight
 			pageRanges := options.PageRanges
 
@@ -67,9 +71,20 @@ func printToPdfActionFunc(reqCtx context.Context, logger *slog.Logger, outputPat
 
 				// There are 96 CSS pixels per inch.
 				// See https://issues.chromium.org/issues/40267771#comment14.
-				// We add top and bottom margins so that the content area
-				// is large enough to fit the entire content.
-				paperHeight = (cssContentSize.Height / 96) + options.MarginTop + options.MarginBottom
+				if options.Landscape {
+					// Landscape swaps the paper dimensions, so the page is
+					// WithPaperHeight wide by WithPaperWidth tall. Size both to
+					// the content so the width expands to fit a wide document
+					// (e.g. a table) instead of the height-only expansion
+					// landing on the width axis and truncating it.
+					// See https://github.com/gotenberg/gotenberg/issues/1390.
+					paperWidth = (cssContentSize.Height / 96) + options.MarginTop + options.MarginBottom
+					paperHeight = (cssContentSize.Width / 96) + options.MarginLeft + options.MarginRight
+				} else {
+					// We add top and bottom margins so that the content area
+					// is large enough to fit the entire content.
+					paperHeight = (cssContentSize.Height / 96) + options.MarginTop + options.MarginBottom
+				}
 				pageRanges = "1" // little dirty hack to avoid leftovers.
 			}
 
@@ -78,7 +93,7 @@ func printToPdfActionFunc(reqCtx context.Context, logger *slog.Logger, outputPat
 				WithLandscape(options.Landscape).
 				WithPrintBackground(options.PrintBackground).
 				WithScale(options.Scale).
-				WithPaperWidth(options.PaperWidth).
+				WithPaperWidth(paperWidth).
 				WithPaperHeight(paperHeight).
 				WithMarginTop(options.MarginTop).
 				WithMarginBottom(options.MarginBottom).
@@ -188,7 +203,16 @@ func captureScreenshotActionFunc(logger *slog.Logger, outputPath string, options
 			WithOptimizeForSpeed(options.OptimizeForSpeed).
 			WithFormat(page.CaptureScreenshotFormat(options.Format))
 
-		if options.Clip {
+		switch {
+		case options.Selector != "":
+			clip, err := elementClip(ctx, options.Selector)
+			if err != nil {
+				return err
+			}
+
+			logger.DebugContext(ctx, fmt.Sprintf("clip screenshot to selector '%s'", options.Selector))
+			captureScreenshot = captureScreenshot.WithClip(clip)
+		case options.Clip:
 			captureScreenshot = captureScreenshot.WithClip(&page.Viewport{
 				Width:  float64(options.Width),
 				Height: float64(options.Height),
@@ -229,6 +253,49 @@ func captureScreenshotActionFunc(logger *slog.Logger, outputPath string, options
 	}
 }
 
+// elementClip resolves the first element matching selector to a page-space clip
+// rectangle for Page.captureScreenshot.
+//
+// getBoundingClientRect reports viewport-relative CSS pixels; adding the scroll
+// offset puts the rectangle in the document coordinate space that
+// WithCaptureBeyondViewport expects. It fails with
+// [ErrScreenshotSelectorNotFound] when nothing matches or the match has no
+// rendered box (display:none or a zero area), so the caller can answer 400.
+func elementClip(ctx context.Context, selector string) (*page.Viewport, error) {
+	var rect struct {
+		Found  bool    `json:"found"`
+		X      float64 `json:"x"`
+		Y      float64 `json:"y"`
+		Width  float64 `json:"width"`
+		Height float64 `json:"height"`
+	}
+
+	expr := fmt.Sprintf(`(() => {
+	const el = document.querySelector(%s);
+	if (!el) {
+		return { found: false };
+	}
+	const r = el.getBoundingClientRect();
+	return { found: true, x: r.left + window.scrollX, y: r.top + window.scrollY, width: r.width, height: r.height };
+})()`, strconv.Quote(selector))
+
+	err := chromedp.Evaluate(expr, &rect).Do(ctx)
+	if err != nil {
+		return nil, fmt.Errorf("evaluate selector box: %v: %w", err, ErrScreenshotSelectorNotFound)
+	}
+	if !rect.Found || rect.Width <= 0 || rect.Height <= 0 {
+		return nil, fmt.Errorf("selector %q matched no element with a visible box: %w", selector, ErrScreenshotSelectorNotFound)
+	}
+
+	return &page.Viewport{
+		X:      rect.X,
+		Y:      rect.Y,
+		Width:  rect.Width,
+		Height: rect.Height,
+		Scale:  1,
+	}, nil
+}
+
 func setDeviceMetricsOverride(logger *slog.Logger, width, height int, deviceScaleFactor float64) chromedp.ActionFunc {
 	return func(ctx context.Context) error {
 		logger.DebugContext(ctx, "set device metrics override")
@@ -280,6 +347,54 @@ func clearCookiesActionFunc(logger *slog.Logger, clear bool) chromedp.ActionFunc
 	}
 }
 
+// clearStorageActionFunc clears the converted origin's local storage before the
+// page loads, so state written by a previous conversion of the same origin does
+// not leak into this one. See https://github.com/gotenberg/gotenberg/issues/919.
+//
+// Session storage is not touched: each conversion runs in its own browsing
+// context (a fresh tab), so it is already isolated and cannot leak. Local
+// storage is per-origin and shared across tabs of the long-lived browser, so it
+// is the only web storage that carries over.
+func clearStorageActionFunc(logger *slog.Logger, clear bool, rawURL string) chromedp.ActionFunc {
+	return func(ctx context.Context) error {
+		if !clear {
+			logger.DebugContext(ctx, "local storage not cleared")
+			return nil
+		}
+
+		origin, ok := httpOrigin(rawURL)
+		if !ok {
+			// A file:// upload gets an opaque, per-request origin that is not
+			// shared between conversions, so there is nothing to clear.
+			logger.DebugContext(ctx, "local storage not cleared: non-http(s) origin is already isolated")
+			return nil
+		}
+
+		logger.DebugContext(ctx, fmt.Sprintf("clear local storage for %s", origin))
+
+		err := storage.ClearDataForOrigin(origin, string(storage.TypeLocalStorage)).Do(ctx)
+		if err == nil {
+			return nil
+		}
+
+		return fmt.Errorf("clear local storage: %w", err)
+	}
+}
+
+// httpOrigin returns the http(s) security origin (scheme://host[:port]) of
+// rawURL, and false when rawURL is not http(s). A non-http(s) URL such as a
+// file:// upload has an opaque origin that no other conversion shares.
+func httpOrigin(rawURL string) (string, bool) {
+	parsed, err := url.Parse(rawURL)
+	if err != nil {
+		return "", false
+	}
+	if parsed.Scheme != "http" && parsed.Scheme != "https" {
+		return "", false
+	}
+	return fmt.Sprintf("%s://%s", parsed.Scheme, parsed.Host), true
+}
+
 func disableJavaScriptActionFunc(logger *slog.Logger, disable bool) chromedp.ActionFunc {
 	return func(ctx context.Context) error {
 		// See https://github.com/gotenberg/gotenberg/issues/175.

test/integration/features/chromium_convert_html.feature

@@ -62,6 +62,24 @@ Feature: /forms/chromium/convert/html
       Page 12
       """
 
+  # A wide table on one landscape page: the page must expand its width to the
+  # content, so the rightmost column is not truncated and the page comes out
+  # landscape instead of the narrow-tall strip the height-only expansion used
+  # to produce. See https://github.com/gotenberg/gotenberg/issues/1390.
+  Scenario: POST /forms/chromium/convert/html (Single Page Landscape)
+    Given I have a default Gotenberg container
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/html" endpoint with the following form data and header(s):
+      | files                     | testdata/wide-table-html/index.html | file   |
+      | singlePage                | true                                | field  |
+      | landscape                 | true                                | field  |
+      | Gotenberg-Output-Filename | foo                                 | header |
+    Then the response status code should be 200
+    Then the response header "Content-Type" should be "application/pdf"
+    Then there should be 1 PDF(s) in the response
+    Then the "foo.pdf" PDF should have 1 page(s)
+    Then the "foo.pdf" PDF should be set to landscape orientation
+    Then the "foo.pdf" PDF should have content matching "Column 12" at page 1
+
   Scenario: POST /forms/chromium/convert/html (Landscape)
     Given I have a default Gotenberg container
     When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/html" endpoint with the following form data and header(s):

test/integration/features/chromium_convert_url.feature

@@ -22,6 +22,41 @@ Feature: /forms/chromium/convert/url
       Page 1
       """
 
+  # localStorage is per-origin and shared by the long-lived browser, so without
+  # clearing it accumulates across same-origin conversions.
+  # See https://github.com/gotenberg/gotenberg/issues/919.
+  Scenario: POST /forms/chromium/convert/url (localStorage leaks without clearing)
+    Given I have a default Gotenberg container
+    Given I have a static server
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/url" endpoint with the following form data and header(s):
+      | url                       | http://host.docker.internal:%d/html/testdata/local-storage-html/index.html | field  |
+      | Gotenberg-Output-Filename | foo                                                                        | header |
+    Then the response status code should be 200
+    Then the "foo.pdf" PDF should have content matching "localStorageCount=1" at page 1
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/url" endpoint with the following form data and header(s):
+      | url                       | http://host.docker.internal:%d/html/testdata/local-storage-html/index.html | field  |
+      | Gotenberg-Output-Filename | foo                                                                        | header |
+    Then the response status code should be 200
+    Then the "foo.pdf" PDF should have content matching "localStorageCount=2" at page 1
+
+  # With --chromium-clear-storage every conversion clears the origin's
+  # localStorage first, so the counter never carries over.
+  # See https://github.com/gotenberg/gotenberg/issues/919.
+  Scenario: POST /forms/chromium/convert/url (Clear Storage)
+    Given I have a Gotenberg container with the following environment variable(s):
+      | CHROMIUM_CLEAR_STORAGE | true |
+    Given I have a static server
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/url" endpoint with the following form data and header(s):
+      | url                       | http://host.docker.internal:%d/html/testdata/local-storage-html/index.html | field  |
+      | Gotenberg-Output-Filename | foo                                                                        | header |
+    Then the response status code should be 200
+    Then the "foo.pdf" PDF should have content matching "localStorageCount=1" at page 1
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/convert/url" endpoint with the following form data and header(s):
+      | url                       | http://host.docker.internal:%d/html/testdata/local-storage-html/index.html | field  |
+      | Gotenberg-Output-Filename | foo                                                                        | header |
+    Then the response status code should be 200
+    Then the "foo.pdf" PDF should have content matching "localStorageCount=1" at page 1
+
   Scenario: POST /forms/chromium/convert/url (Single Page)
     Given I have a default Gotenberg container
     Given I have a static server

test/integration/features/chromium_screenshot_html.feature

@@ -69,6 +69,32 @@ Feature: /forms/chromium/screenshot/html
     Then the response status code should be 200
     Then the response header "Content-Type" should be "image/png"
 
+  # The target element is 300x200 and sits below a spacer, so a correct clip
+  # proves both the element size and its page offset. See issue #947.
+  Scenario: POST /forms/chromium/screenshot/html (Selector)
+    Given I have a default Gotenberg container
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/screenshot/html" endpoint with the following form data and header(s):
+      | files                     | testdata/screenshot-selector-html/index.html | file   |
+      | selector                  | #target                                      | field  |
+      | Gotenberg-Output-Filename | foo                                          | header |
+    Then the response status code should be 200
+    Then the response header "Content-Type" should be "image/png"
+    Then the "foo.png" image should be 300x200 pixels
+    # A centered red pixel proves the clip landed on the element, not on the
+    # white spacer above it, i.e. the page offset was applied.
+    Then the "foo.png" image pixel at 150,100 should be "#ff0000"
+
+  Scenario: POST /forms/chromium/screenshot/html (Selector Not Found)
+    Given I have a default Gotenberg container
+    When I make a "POST" request to Gotenberg at the "/forms/chromium/screenshot/html" endpoint with the following form data and header(s):
+      | files    | testdata/screenshot-selector-html/index.html | file  |
+      | selector | #does-not-exist                              | field |
+    Then the response status code should be 400
+    Then the response body should contain string:
+      """
+      The selector '#does-not-exist' (selector) matched no element with a visible box
+      """
+
   Scenario: POST /forms/chromium/screenshot/html (Quality)
     Given I have a default Gotenberg container
     When I make a "POST" request to Gotenberg at the "/forms/chromium/screenshot/html" endpoint with the following form data and header(s):

test/integration/features/debug.feature

@@ -83,6 +83,7 @@ Feature: /debug
           "chromium-auto-start": "false",
           "chromium-clear-cache": "false",
           "chromium-clear-cookies": "false",
+          "chromium-clear-storage": "false",
           "chromium-deny-list": "[^file:(?!//\\/tmp/).*]",
           "chromium-deny-private-ips": "false",
           "chromium-deny-public-ips": "false",
@@ -221,6 +222,7 @@ Feature: /debug
           "chromium-auto-start": "false",
           "chromium-clear-cache": "false",
           "chromium-clear-cookies": "false",
+          "chromium-clear-storage": "false",
           "chromium-deny-list": "[^file:(?!//\\/tmp/).*]",
           "chromium-deny-private-ips": "false",
           "chromium-deny-public-ips": "false",

test/integration/scenario/scenario.go

@@ -5,6 +5,8 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"image"
+	_ "image/png" // Register the PNG decoder for image.DecodeConfig.
 	"io"
 	"mime"
 	"net/http"
@@ -1010,6 +1012,51 @@ func (s *scenario) thePdfShouldHaveImages(ctx context.Context, name string, imag
 	return nil
 }
 
+func (s *scenario) theImageShouldBePixels(_ context.Context, name string, width, height int) error {
+	path := fmt.Sprintf("%s/%s/%s", s.workdir, s.resp.Header().Get("Gotenberg-Trace"), name)
+
+	file, err := os.Open(path) //nolint:gosec // path is built from test-controlled values.
+	if err != nil {
+		return fmt.Errorf("open image %q: %w", path, err)
+	}
+	defer file.Close()
+
+	config, format, err := image.DecodeConfig(file)
+	if err != nil {
+		return fmt.Errorf("decode image %q: %w", path, err)
+	}
+
+	if config.Width != width || config.Height != height {
+		return fmt.Errorf("expected %s image %dx%d, but actual is %dx%d", format, width, height, config.Width, config.Height)
+	}
+
+	return nil
+}
+
+func (s *scenario) theImagePixelShouldBe(_ context.Context, name string, x, y int, want string) error {
+	path := fmt.Sprintf("%s/%s/%s", s.workdir, s.resp.Header().Get("Gotenberg-Trace"), name)
+
+	file, err := os.Open(path) //nolint:gosec // path is built from test-controlled values.
+	if err != nil {
+		return fmt.Errorf("open image %q: %w", path, err)
+	}
+	defer file.Close()
+
+	img, format, err := image.Decode(file)
+	if err != nil {
+		return fmt.Errorf("decode image %q: %w", path, err)
+	}
+
+	r, g, b, _ := img.At(x, y).RGBA()
+	// RGBA returns 16-bit channels; shift down to the 8-bit hex form.
+	got := fmt.Sprintf("#%02x%02x%02x", r>>8, g>>8, b>>8)
+	if !strings.EqualFold(got, want) {
+		return fmt.Errorf("expected %s pixel at %d,%d to be %s, but actual is %s", format, x, y, want, got)
+	}
+
+	return nil
+}
+
 func (s *scenario) thePdfShouldBeSetToLandscapeOrientation(ctx context.Context, name string, kind string) error {
 	var path string
 	if !strings.HasPrefix(name, "*_") {
@@ -1047,7 +1094,9 @@ func (s *scenario) thePdfShouldBeSetToLandscapeOrientation(ctx context.Context,
 	}
 
 	output = strings.ReplaceAll(output, " ", "")
-	re := regexp.MustCompile(`Pagesize:(\d+)x(\d+).*`)
+	// The dimensions can be fractional (e.g. a content-sized single page),
+	// so match floats, not just integers.
+	re := regexp.MustCompile(`Pagesize:([\d.]+)x([\d.]+)`)
 	matches := re.FindStringSubmatch(output)
 
 	if len(matches) < 3 {
@@ -1056,22 +1105,22 @@ func (s *scenario) thePdfShouldBeSetToLandscapeOrientation(ctx context.Context,
 
 	invert := kind == "should NOT"
 
-	width, err := strconv.Atoi(matches[1])
+	width, err := strconv.ParseFloat(matches[1], 64)
 	if err != nil {
-		return fmt.Errorf("convert width value %q to integer: %w", matches[1], err)
+		return fmt.Errorf("convert width value %q to float: %w", matches[1], err)
 	}
 
-	height, err := strconv.Atoi(matches[2])
+	height, err := strconv.ParseFloat(matches[2], 64)
 	if err != nil {
-		return fmt.Errorf("convert height value %q to integer: %w", matches[2], err)
+		return fmt.Errorf("convert height value %q to float: %w", matches[2], err)
 	}
 
 	if invert && height < width {
-		return fmt.Errorf("expected height %d to be greater than width %d", height, width)
+		return fmt.Errorf("expected height %g to be greater than width %g", height, width)
 	}
 
 	if !invert && width < height {
-		return fmt.Errorf("expected width %d to be greater than height %d", width, height)
+		return fmt.Errorf("expected width %g to be greater than height %g", width, height)
 	}
 
 	return nil
@@ -1599,6 +1648,8 @@ func InitializeScenario(ctx *godog.ScenarioContext) {
 	ctx.Then(`^the "([^"]*)" PDF (should|should NOT) have the following content at page (\d+):$`, s.thePdfShouldHaveTheFollowingContentAtPage)
 	ctx.Then(`^the "([^"]*)" PDF (should|should NOT) have content matching "([^"]*)" at page (\d+)$`, s.thePdfShouldHaveContentMatchingAtPage)
 	ctx.Then(`^the "([^"]*)" PDF should have (\d+) image\(s\)$`, s.thePdfShouldHaveImages)
+	ctx.Then(`^the "([^"]*)" image should be (\d+)x(\d+) pixels$`, s.theImageShouldBePixels)
+	ctx.Then(`^the "([^"]*)" image pixel at (\d+),(\d+) should be "([^"]*)"$`, s.theImagePixelShouldBe)
 	ctx.After(func(ctx context.Context, sc *godog.Scenario, err error) (context.Context, error) {
 		if s.gotenbergContainer != nil {
 			errTerminate := s.gotenbergContainer.Terminate(ctx, testcontainers.StopTimeout(0))

test/integration/testdata/local-storage-html/index.html

@@ -0,0 +1,18 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <title>Local storage counter</title>
+  </head>
+  <body>
+    <!-- Increments a per-origin localStorage counter and renders it, so a value
+         above 1 means a previous conversion of the same origin leaked into this
+         one. See https://github.com/gotenberg/gotenberg/issues/919. -->
+    <pre id="out"></pre>
+    <script>
+      var n = (parseInt(localStorage.getItem("n"), 10) || 0) + 1;
+      localStorage.setItem("n", n);
+      document.getElementById("out").textContent = "localStorageCount=" + n;
+    </script>
+  </body>
+</html>

test/integration/testdata/screenshot-selector-html/index.html

@@ -0,0 +1,28 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <title>Element screenshot</title>
+    <style>
+      body {
+        margin: 0;
+        padding: 40px;
+        background: #ffffff;
+      }
+      /* A spacer pushes the target away from the origin so the clip has to
+         account for the element offset, not just its size. */
+      .spacer {
+        height: 120px;
+      }
+      #target {
+        width: 300px;
+        height: 200px;
+        background: #ff0000;
+      }
+    </style>
+  </head>
+  <body>
+    <div class="spacer"></div>
+    <div id="target"></div>
+  </body>
+</html>

test/integration/testdata/wide-table-html/index.html

@@ -0,0 +1,43 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <title>Wide table</title>
+    <style>
+      html,
+      body {
+        margin: 0;
+        padding: 0;
+      }
+      table {
+        border-collapse: collapse;
+        white-space: nowrap;
+      }
+      td {
+        border: 1px solid #000;
+        padding: 8px 24px;
+      }
+    </style>
+  </head>
+  <body>
+    <!-- A single wide row: much wider than tall, so a correct single page is
+         landscape, and the rightmost column only survives if the width was
+         expanded to the content instead of truncated. -->
+    <table>
+      <tr>
+        <td>Column 1</td>
+        <td>Column 2</td>
+        <td>Column 3</td>
+        <td>Column 4</td>
+        <td>Column 5</td>
+        <td>Column 6</td>
+        <td>Column 7</td>
+        <td>Column 8</td>
+        <td>Column 9</td>
+        <td>Column 10</td>
+        <td>Column 11</td>
+        <td>Column 12</td>
+      </tr>
+    </table>
+  </body>
+</html>